Soccer predictions: actual data vs predicted data

Slate 20261018 · 8 match(es)

How to read this report. Each match shows the model's headline prediction, the stage-by-stage lambda (goal-rate) build, both teams' standing/form inputs, and every starter's projected line next to their real season baseline. A moderate / large deviation badge flags a projection that strays far from the player's season form — a quick way to spot when the inputs look right but a projection does not. When the real result is known an actual-vs-predicted banner appears.

All matches overview

Match	Score	Winner	H / D / A	O2.5	BTTS	Conf	λ H/A	Actual	Result
Celta vs Alavés LaLiga	0-0	Draw	33.1% / 34.3% / 32.6%	24.5%	33.1%	0.69	0.86 / 0.85	-	pending
Forest vs Arsenal Premier League	1-1	Draw	23.5% / 23.3% / 53.1%	55.6%	56.1%	0.75	1.12 / 1.79	-	pending
Brighton vs Crystal Palace Premier League	1-1	Draw	41.9% / 26.2% / 31.9%	49.3%	53.6%	0.7	1.43 / 1.22	-	pending
Leeds vs Man Utd Premier League	1-1	Draw	30.0% / 25.5% / 44.5%	51.4%	54.9%	0.72	1.21 / 1.52	-	pending
Málaga vs Real Sociedad LaLiga	0-1	Real Sociedad	16.0% / 26.2% / 57.9%	36.5%	37.2%	0.76	0.65 / 1.50	-	pending
Osasuna vs Racing Club LaLiga	1-1	Draw	35.0% / 26.1% / 38.9%	50.5%	54.9%	0.7	1.30 / 1.39	-	pending
Real Madrid vs Sevilla LaLiga	3-1	Real Madrid	79.0% / 11.4% / 9.6%	86.3%	70.7%	0.82	3.57 / 1.30	-	pending
Bournemouth vs Sunderland Premier League	1-1	Draw	43.0% / 28.5% / 28.5%	39.5%	45.4%	0.71	1.28 / 0.99	-	pending

Match detail

Celta vs Alavés

LaLiga · 20261018 · 3-4-3 vs 5-3-2

Predicted 0-0Winner: DrawWinning H 1-0 (15.4%)Winning A 0-1 (15.2%)Confidence 0.69λ home/away 0.86 / 0.85Most-likely P 18.3%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Celta vs Alavés. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 3-4-3 vs 5-3-2 (duel lambdas: Home 0.835, Away 0.874). Final calibrated Poisson lambdas: Home 0.859, Away 0.849 (most likely score 0-0).

Home Starters 11/11 XGBoost (9 FBref) Away Starters 11/11 XGBoost (9 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 0.83 / 0.87 Final λ (H/A) 0.86 / 0.85

FBref Team Priors: Celta Vigo (Gls/90: 0.25, GA/90: 0.25 → E[λ]: 0.309) vs Deportivo Alavés (Gls/90: 1.6666666666666667, GA/90: 1.67 → E[λ]: 0.309) | Blend Weight: 0.228

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Celta	Alavés
League position	9.00	7.00
Points per game	1.16	1.37
Attack index	0.86	0.99
Defense (concede) index	0.86	0.79
Form multiplier	0.99	1.01
Standing λ multiplier	0.88	0.91

Lambda decomposition (how the goal rates were built)

Stage	Celta λ	Alavés λ
FBref Squad Prior E[λ] (weight 0.228)	0.309	0.309
Pre-competitive-floor	0.859	0.849
Pre-calibration (Micro-duels)	0.835	0.874
Final λ	0.859	0.849

competition calibration scale 1.0competition observed total goals 0.0quality gap applied Falsemismatch score pressure applied Falsecompetitive underdog floor applied Falsestage discount 1.0

Scoreline distribution & markets

Top 5 scorelines	P
0-0	18.3%
1-0	15.4%
0-1	15.2%
1-1	13.4%
2-0	6.7%

Home win	33.1%
Draw	34.3%
Away win	32.6%
Over 2.5	24.5%
BTTS yes	33.1%

Pregame fair line — favorite Celta (33.1%), fair total 1.71, fair spread 0.01

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Celta — projected line vs season baseline (3-4-3)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Hugo Álvarez F 73 0 0.15 0.08 85% 1 0.4 6.8 53.7 38 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9997, quality 0.496)	`market_value_eur=7800000` `quality_factor=0.4963` `boost_factor=0.9997` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0072)	`blend_weight=0.0072` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.8062` `prior_reliability=0.0867` `competition=La Liga`	totalCross: 0.9998 -> 1.0195 totalShots: 1.2189 -> 1.2101
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ville Koski', 'matchup_zone': 'Shadow Striker vs Centre Back — Half-Space Threat', 'duel_win_rate': 0.425, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9637, 'skill_edge': 0.0113, 'height_edge': -0.036, 'attacker_skill': 0.4963, 'defender_skill': 0.4712, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ville Koski', 'market_value': 4300000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4712}], 'xg_override': 0.1324}	totalShots: 1.2158 -> 1.1717 shotOffTarget: 0.7951 -> 0.7662 onTargetScoringAttempt: 0.4207 -> 0.4054 goals: 0.1562 -> 0.1505 bigChanceMissed: 0.1437 -> 0.1385 expectedGoals: 0.1374 -> 0.1324
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_goals: 0.1324 -> 0.1484 expected_assists: 0.0764 -> 0.0856
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 75.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.1379` `goal_rate=0.1505` `scorer_weight=0.1631`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.082` `assist_rate=0.3603` `provider_weight=0.107`	-

Iago Aspas F 67 0 0.16 0.08 105% 2 0.7 6.8 54.1 29 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9976, quality 0.437)	`market_value_eur=1500000` `quality_factor=0.4374` `boost_factor=0.9976` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0086)	`blend_weight=0.0086` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=0.7443` `prior_reliability=0.1133` `competition=La Liga`	totalCross: 0.7874 -> 0.7807 wasFouled: 1.0956 -> 1.09
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ville Koski, Nahuel Tenaglia', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9395, 'skill_edge': -0.0111, 'height_edge': -0.03, 'attacker_skill': 0.4374, 'defender_skill': 0.4622, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ville Koski', 'market_value': 4300000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4712}, {'player': 'Nahuel Tenaglia', 'market_value': 3200000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4531}], 'xg_override': 0.145}	totalShots: 1.4304 -> 1.3439 shotOffTarget: 0.8924 -> 0.8384 onTargetScoringAttempt: 0.538 -> 0.5055 expectedGoalsOnTarget: 0.2097 -> 0.197 goals: 0.1785 -> 0.1677 expectedGoals: 0.1543 -> 0.145
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_goals: 0.145 -> 0.1625 expected_assists: 0.0797 -> 0.0893
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 81.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.15` `goal_rate=0.1677` `scorer_weight=0.1814`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0854` `assist_rate=0.0` `provider_weight=0.0714`	-

Ferrán Jutglà F 73 0 0.21 0.17 ok 2 1.2 6.7 59.4 28 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0009, quality 0.531)	`market_value_eur=10700000` `quality_factor=0.5306` `boost_factor=1.0009` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0131)	`blend_weight=0.0131` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=0.8092` `prior_reliability=0.1733` `competition=La Liga`	totalShots: 1.568 -> 1.5516 wasFouled: 1.5501 -> 1.5583 onTargetScoringAttempt: 0.6078 -> 0.5998 totalOffside: 0.5433 -> 0.5362
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ville Koski', 'matchup_zone': 'Shadow Striker vs Centre Back — Half-Space Threat', 'duel_win_rate': 0.437, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9805, 'skill_edge': 0.0268, 'height_edge': -0.04, 'attacker_skill': 0.5306, 'defender_skill': 0.4712, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ville Koski', 'market_value': 4300000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4712}], 'xg_override': 0.1884}	totalShots: 1.56 -> 1.5296 shotOffTarget: 0.9602 -> 0.9415 onTargetScoringAttempt: 0.5998 -> 0.5881 expectedGoalsOnTarget: 0.2996 -> 0.2938
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_goals: 0.1884 -> 0.2113 expected_assists: 0.0537 -> 0.0602
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 16 by projected rating, fifa_overall 75.0)	`match_squad_rank=16` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.1905` `goal_rate=0.2025` `scorer_weight=0.2238`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0584` `assist_rate=0.0728` `provider_weight=0.0713`	-

Javi Galán D 77 0 0.01 0.01 ok 0 0.2 6.8 60.4 34 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9973, quality 0.437)	`market_value_eur=1700000` `quality_factor=0.4374` `boost_factor=0.9973` `reference_quality=0.505` `history_matches=32` `history_shrink=0.1351`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=32` `season_overlap_discount=1.0` `per90_to_match_factor=0.8568` `prior_reliability=0.1` `competition=La Liga`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Toni Martínez', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_assists: 0.0682 -> 0.0764
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 77.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0056` `goal_rate=0.0` `scorer_weight=0.0051`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0736` `assist_rate=0.0058` `provider_weight=0.0611`	-

Hugo Burcio M 86 0 0.10 0.03 272% 1 0.2 6.7 22.1 12 XGBoost M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9887, quality 0.437)	`market_value_eur=1900000` `quality_factor=0.4374` `boost_factor=0.9887` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ville Koski', 'matchup_zone': 'LWB vs CB', 'duel_win_rate': 0.413, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9894, 'skill_edge': -0.0152, 'height_edge': 0.008, 'attacker_skill': 0.4374, 'defender_skill': 0.4712, 'attacker_role': 'LWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Ville Koski', 'market_value': 4300000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4712}], 'xg_override': 0.0849}`	aerialWon: 0.2241 -> 1.2 totalShots: 1.0343 -> 1.0233 shotOffTarget: 0.7258 -> 0.7181
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_goals: 0.0849 -> 0.0951 expected_assists: 0.0443 -> 0.0496
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 15 by projected rating, fifa_overall 67.0)	`match_squad_rank=15` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0907` `goal_rate=0.1465` `scorer_weight=0.1189`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0484` `assist_rate=0.0393` `provider_weight=0.047`	-

Aleix Febas M 85 0 0.08 0.04 114% 1 0.5 6.9 59.7 26 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9986, quality 0.465)	`market_value_eur=5200000` `quality_factor=0.4646` `boost_factor=0.9986` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0089)	`blend_weight=0.0089` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9408` `prior_reliability=0.12` `competition=La Liga`	fouls: 1.5694 -> 1.5787
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Pablo Ibáñez', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.411, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9431, 'skill_edge': 0.0053, 'height_edge': -0.044, 'attacker_skill': 0.4717, 'defender_skill': 0.46, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Pablo Ibáñez', 'market_value': 4099999, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.46}], 'xg_override': 0.0692}	aerialWon: 0.109 -> 0.8 totalShots: 0.811 -> 0.7649 shotOffTarget: 0.4816 -> 0.4542 onTargetScoringAttempt: 0.3294 -> 0.3107
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_goals: 0.0692 -> 0.0775 expected_assists: 0.0842 -> 0.0944
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 79.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0746` `goal_rate=0.068` `scorer_weight=0.0767`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0901` `assist_rate=0.0689` `provider_weight=0.0836`	-

Álvaro Núñez D 81 0 0.02 0.01 160% 1 0.3 6.8 34.3 26 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9975, quality 0.449)	`market_value_eur=3900000` `quality_factor=0.4486` `boost_factor=0.9975` `reference_quality=0.505` `history_matches=29` `history_shrink=0.1471`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0121)	`blend_weight=0.0121` `history_matches=29` `season_overlap_discount=1.0` `per90_to_match_factor=0.9022` `prior_reliability=0.1267` `competition=La Liga`	interceptionWon: 0.7721 -> 0.78 totalCross: 1.49 -> 1.4835 wonTackle: 0.9129 -> 0.9076
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Lucas Boyé', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	expected_assists: 0.0588 -> 0.0659
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 76.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0156` `goal_rate=0.0` `scorer_weight=0.0126`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0638` `assist_rate=0.0204` `provider_weight=0.0527`	-

Yoel Lago D 89 0 0.02 0.00 236% 0 0.1 6.8 42.1 26 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9989, quality 0.487)	`market_value_eur=7000000` `quality_factor=0.4866` `boost_factor=0.9989` `reference_quality=0.505` `history_matches=20` `history_shrink=0.2`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0303)	`blend_weight=0.0303` `history_matches=20` `season_overlap_discount=1.0` `per90_to_match_factor=0.9912` `prior_reliability=0.2333` `competition=La Liga`	fouls: 2.1519 -> 2.1382 totalShots: 0.3629 -> 0.3519 totalCross: 0.0 -> 0.0086 interceptionWon: 1.1822 -> 1.1893 wonTackle: 0.7893 -> 0.7825
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jonny Otto', 'matchup_zone': 'LWB vs CB'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 71.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0151` `goal_rate=0.0` `scorer_weight=0.0122`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0129` `assist_rate=0.0` `provider_weight=0.0106`	-

Sebastián Cáceres D 90 0 0.01 0.00 206% 1 0.2 6.8 23.8 6 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9884, quality 0.451)	`market_value_eur=4099999` `quality_factor=0.4511` `boost_factor=0.9884` `reference_quality=0.505` `history_matches=2` `history_shrink=0.7143`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Lucas Boyé', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 75.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0133` `goal_rate=0.0` `scorer_weight=0.0108`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0129` `assist_rate=0.0` `provider_weight=0.0106`	-

Ionuț Radu G 90 0 0.00 - n/a 0.0 6.9 65.0 31 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0004, quality 0.519)	`market_value_eur=9700000` `quality_factor=0.5189` `boost_factor=1.0004` `reference_quality=0.505` `history_matches=46` `history_shrink=0.098`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0170)	`blend_weight=0.017` `history_matches=46` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2667` `competition=La Liga`	saves: 2.882 -> 2.8925
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Toni Martínez, Lucas Boyé', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 0.8739}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	shots_faced: 8.3 -> 8.1 expected_saves: 2.03 -> 1.99 high_pressure_saves: 0.7 -> 0.68
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 78.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=0.9142`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.6035`	-

Marcos Alonso D 87 0 0.04 0.08 45% 1 0.7 7.1 62.2 27 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9976, quality 0.437)	`market_value_eur=1300000` `quality_factor=0.4374` `boost_factor=0.9976` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0131)	`blend_weight=0.0131` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=0.963` `prior_reliability=0.1733` `competition=La Liga`	interceptionWon: 1.1502 -> 1.1739 wonTackle: 0.9483 -> 0.9553
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Abderrahman Rebbach', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.86 (squad xG sum 0.7665, xG scale 1.1206, xA scale 1.1206, team xA 0.64 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=0.859` `squad_xg_sum=0.7665` `squad_xa_sum=0.5741` `team_expected_assists=0.6433` `team_xa_to_xg_ratio=0.7489`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 79.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0407` `goal_rate=0.0061` `scorer_weight=0.0332`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0453` `assist_rate=0.0` `provider_weight=0.0371`	-

Team totals 0 0.86 11 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Altay Bayındır	G	4.5m	11.3%	Market-value substitute prior
Javi Rodríguez	D	10.3m	32.7%	Market-value substitute prior
Abdoulaye Faye	D	13.6m	21.8%	Market-value substitute prior
Fer López	M	22.6m	52.9%	Market-value substitute prior
Miguel Román	M	34.9m	41.4%	Market-value substitute prior
Williot Swedberg	F	30.4m	44.5%	Market-value substitute prior

Alavés — projected line vs season baseline (5-3-2)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Ville Koski D 89 0 0.09 0.04 120% 0 0.1 6.7 37.1 22 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9957, quality 0.454)	`market_value_eur=4300000` `quality_factor=0.4535` `boost_factor=0.9957` `reference_quality=0.505` `history_matches=13` `history_shrink=0.2778`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0349)	`blend_weight=0.0349` `history_matches=13` `season_overlap_discount=1.0` `per90_to_match_factor=0.9923` `prior_reliability=0.1933` `competition=La Liga`	fouls: 1.3379 -> 1.2912 totalShots: 0.3872 -> 0.3737 onTargetScoringAttempt: 0.3521 -> 0.3398 wonTackle: 0.6428 -> 0.6323 goals: 0.1652 -> 0.1594 wasFouled: 0.1581 -> 0.1526
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Hugo Burcio', 'matchup_zone': 'LWB vs CB'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1267 -> 0.0931
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 14 by projected rating, fifa_overall 68.0)	`match_squad_rank=14` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0889` `goal_rate=0.1594` `scorer_weight=0.1103`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0159` `assist_rate=0.0` `provider_weight=0.0128`	-

Jonny Otto D 88 0 0.01 0.01 ok 0 0.2 6.8 53.6 30 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9978, quality 0.437)	`market_value_eur=2100000` `quality_factor=0.4374` `boost_factor=0.9978` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Yoel Lago', 'matchup_zone': 'LWB vs CB'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 84.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0107` `goal_rate=0.0` `scorer_weight=0.0086`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0299` `assist_rate=0.0` `provider_weight=0.0243`	-

Nahuel Tenaglia D 86 0 0.03 0.04 36% 1 0.6 6.8 61.5 31 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9980, quality 0.440)	`market_value_eur=3200000` `quality_factor=0.4399` `boost_factor=0.998` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0133)	`blend_weight=0.0133` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=0.9558` `prior_reliability=0.2` `competition=La Liga`	totalShots: 0.6318 -> 0.6403 fouls: 0.7834 -> 0.7899 interceptionWon: 1.1823 -> 1.1877 wasFouled: 1.3585 -> 1.3532
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Iago Aspas', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0354 -> 0.026
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 77.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0257` `goal_rate=0.0829` `scorer_weight=0.0325`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0351` `assist_rate=0.0` `provider_weight=0.0286`	-

Antonio Sivera G 90 0 0.00 - n/a 0.0 7.0 65.3 31 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9990, quality 0.470)	`market_value_eur=5600000` `quality_factor=0.4695` `boost_factor=0.999` `reference_quality=0.505` `history_matches=46` `history_shrink=0.098`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0127)	`blend_weight=0.0127` `history_matches=46` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2` `competition=La Liga`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ferrán Jutglà, Iago Aspas', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 0.8347}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	shots_faced: 7.9 -> 8.2 expected_saves: 1.93 -> 2.01 high_pressure_saves: 0.67 -> 0.69
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 85.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=0.9123`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.5324`	-

Nicolás Valentini D 86 0 0.00 0.00 100% 0 0.1 6.7 38.4 9 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9901, quality 0.439)	`market_value_eur=3100000` `quality_factor=0.4387` `boost_factor=0.9901` `reference_quality=0.505` `history_matches=5` `history_shrink=0.5`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0022)	`blend_weight=0.0022` `history_matches=5` `season_overlap_discount=1.0` `per90_to_match_factor=0.9508` `prior_reliability=0.0067` `competition=La Liga`	fouls: 0.3849 -> 0.4048 totalShots: 0.3776 -> 0.3937
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ferrán Jutglà', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 73.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0` `goal_rate=0.0` `scorer_weight=0.001`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.023` `assist_rate=0.0` `provider_weight=0.0186`	-

Abderrahman Rebbach F 80 0 0.05 0.04 38% 1 0.6 6.8 51.0 41 XGBoost F_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9976, quality 0.437)	`market_value_eur=2400000` `quality_factor=0.4374` `boost_factor=0.9976` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yoel Lago, Sebastián Cáceres', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9174, 'skill_edge': -0.0242, 'height_edge': -0.032, 'attacker_skill': 0.4374, 'defender_skill': 0.4913, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Yoel Lago', 'market_value': 7000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5146}, {'player': 'Sebastián Cáceres', 'market_value': 4099999, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4679}], 'xg_override': 0.0745}	totalShots: 0.9894 -> 0.9077 shotOffTarget: 0.6806 -> 0.6244 onTargetScoringAttempt: 0.3088 -> 0.2833 goals: 0.0927 -> 0.085 expectedGoals: 0.0812 -> 0.0745
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0745 -> 0.0547
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 78.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0532` `goal_rate=0.085` `scorer_weight=0.0593`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0485` `assist_rate=0.0758` `provider_weight=0.0621`	-

Antonio Blanco M 83 0 0.04 0.04 ok 1 1.1 6.8 64.0 31 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0004, quality 0.519)	`market_value_eur=9700000` `quality_factor=0.5189` `boost_factor=1.0004` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0133)	`blend_weight=0.0133` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=0.9168` `prior_reliability=0.2` `competition=La Liga`	fouls: 1.5327 -> 1.5489 interceptionWon: 1.4305 -> 1.444 totalCross: 0.9861 -> 0.9933 totalShots: 0.792 -> 0.7978
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Aleix Febas', 'matchup_zone': 'Central Mid vs Central Mid — Box-to-Box Midfield Contest', 'duel_win_rate': 0.479, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.0866, 'skill_edge': 0.0269, 'height_edge': 0.032, 'attacker_skill': 0.5316, 'defender_skill': 0.4717, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Aleix Febas', 'market_value': 5200000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.4717}], 'xg_override': 0.0591}	aerialWon: 0.281 -> 0.9 totalShots: 0.7978 -> 0.8669 shotOffTarget: 0.5805 -> 0.6308 onTargetScoringAttempt: 0.2086 -> 0.2267
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0591 -> 0.0434
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 78.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0425` `goal_rate=0.0606` `scorer_weight=0.044`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0533` `assist_rate=0.0758` `provider_weight=0.0674`	-

Denis Suárez M 72 0 0.07 0.03 117% 1 0.3 6.9 54.6 27 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9976, quality 0.437)	`market_value_eur=1100000` `quality_factor=0.4374` `boost_factor=0.9976` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0062)	`blend_weight=0.0062` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.7968` `prior_reliability=0.08` `competition=La Liga`	totalCross: 3.7138 -> 3.7484 fouls: 0.6893 -> 0.6974
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Aleix Febas', 'matchup_zone': 'Central Mid vs Central Mid — Box-to-Box Midfield Contest', 'duel_win_rate': 0.437, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.0244, 'skill_edge': -0.0154, 'height_edge': 0.032, 'attacker_skill': 0.4374, 'defender_skill': 0.4717, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Aleix Febas', 'market_value': 5200000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.4717}], 'xg_override': 0.0886}	aerialWon: 0.1102 -> 0.9 totalShots: 0.9436 -> 0.9666 shotOffTarget: 0.6312 -> 0.6466 onTargetScoringAttempt: 0.3124 -> 0.32
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0886 -> 0.0651 expected_assists: 0.1605 -> 0.1683
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 74.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.063` `goal_rate=0.0994` `scorer_weight=0.0702`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.1549` `assist_rate=0.1369` `provider_weight=0.1494`	-

Lucas Boyé F 76 0 0.21 0.34 39% 2 2.0 6.7 61.4 26 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9986, quality 0.466)	`market_value_eur=5300000` `quality_factor=0.4659` `boost_factor=0.9986` `reference_quality=0.505` `history_matches=36` `history_shrink=0.122`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0085)	`blend_weight=0.0085` `history_matches=36` `season_overlap_discount=1.0` `per90_to_match_factor=0.8422` `prior_reliability=0.1067` `competition=La Liga`	wonTackle: 1.1926 -> 1.2048 wasFouled: 1.8194 -> 1.8129
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yoel Lago, Sebastián Cáceres', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.185, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9773, 'skill_edge': -0.0114, 'height_edge': -0.004, 'attacker_skill': 0.4659, 'defender_skill': 0.4913, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Yoel Lago', 'market_value': 7000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5146}, {'player': 'Sebastián Cáceres', 'market_value': 4099999, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4679}], 'xg_override': 0.2833}	totalShots: 2.1417 -> 2.0931 shotOffTarget: 1.4168 -> 1.3846 onTargetScoringAttempt: 0.7234 -> 0.707 expectedGoalsOnTarget: 0.3956 -> 0.3866 expectedGoals: 0.2899 -> 0.2833 goals: 0.2834 -> 0.277
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.2833 -> 0.2082
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 77.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.188` `goal_rate=0.277` `scorer_weight=0.2043`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0437` `assist_rate=0.008` `provider_weight=0.0358`	-

Toni Martínez F 77 0 0.24 0.34 ok 2 2.6 6.8 66.0 30 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9995, quality 0.490)	`market_value_eur=7300000` `quality_factor=0.4903` `boost_factor=0.9995` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0115)	`blend_weight=0.0115` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=0.8537` `prior_reliability=0.1733` `competition=La Liga`	onTargetScoringAttempt: 0.8653 -> 0.8554
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yoel Lago, Sebastián Cáceres', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.184, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9758, 'skill_edge': -0.0005, 'height_edge': -0.016, 'attacker_skill': 0.4903, 'defender_skill': 0.4913, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Yoel Lago', 'market_value': 7000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5146}, {'player': 'Sebastián Cáceres', 'market_value': 4099999, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4679}], 'xg_override': 0.33}	totalShots: 2.329 -> 2.2726 shotOffTarget: 1.464 -> 1.4286 onTargetScoringAttempt: 0.8554 -> 0.8347 goals: 0.3896 -> 0.3802 bigChanceMissed: 0.3881 -> 0.3787 expectedGoals: 0.3382 -> 0.33 expectedGoalsOnTarget: 0.2673 -> 0.2608
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.33 -> 0.2426
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 78.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.2154` `goal_rate=0.3802` `scorer_weight=0.2746`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0342` `assist_rate=0.0209` `provider_weight=0.0297`	-

Pablo Ibáñez M 80 0 0.05 0.08 36% 1 1.2 6.8 63.1 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9984, quality 0.451)	`market_value_eur=4099999` `quality_factor=0.4511` `boost_factor=0.9984` `reference_quality=0.505` `history_matches=45` `history_shrink=0.1`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0130)	`blend_weight=0.013` `history_matches=45` `season_overlap_discount=1.0` `per90_to_match_factor=0.8892` `prior_reliability=0.2` `competition=La Liga`	fouls: 1.7146 -> 1.7077 totalShots: 0.9808 -> 0.9874 interceptionWon: 1.0629 -> 1.0683
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Aleix Febas', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.85 (squad xG sum 1.1554, xG scale 0.7348, xA scale 1.0488, team xA 0.55 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=0.849` `squad_xg_sum=1.1554` `squad_xa_sum=0.5262` `team_expected_assists=0.5518` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0736 -> 0.0541
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 73.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0527` `goal_rate=0.0611` `scorer_weight=0.0486`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.046` `assist_rate=0.0775` `provider_weight=0.0589`	-

Team totals 0 0.85 10 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Grégoire Swiderski	G	4.4m	10.9%	Market-value substitute prior
Youssef Enriquez	D	13.7m	30.4%	Market-value substitute prior
Victor Parada	D	13.6m	30.1%	Market-value substitute prior
Ángel Pérez	M	22.1m	30.9%	Market-value substitute prior
Carles Aleñá	M	13.2m	69.2%	Market-value substitute prior
Miguel Rodríguez	F	25.4m	56.2%	Market-value substitute prior

Antagonist / tactical summary

Midfield is evenly contested — Aleix Febas vs Antonio Blanco will be the pivotal duel. Total antagonist xG — Home: 0.62, Away: 0.85. Home danger zone: Shadow Striker vs Centre Back — Half-Space Threat. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Midfield — possession 49.3% / 50.7%, dominance ratio 0.48, key duel: Aleix Febas vs Antonio Blanco

Hugo Burcio xG 0.08Aleix Febas xG 0.07Hugo Álvarez xG 0.13Jonny Otto xG 0.01Antonio Blanco xG 0.06Denis Suárez xG 0.09

Summary

Antagonist matchup analysis for Celta vs Alavés. Most likely scoreline: 0-0. Lambdas derived from 5 home and 6 away individual player duels. Head-to-head: Celta 6W-2D-2L vs Alavés over 10 meetings.

Tactical parity between 3-4-3 and 5-3-2. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Aleix Febas vs Antonio Blanco will be the pivotal duel. Total antagonist xG — Home: 0.62, Away: 0.85. Home danger zone: Shadow Striker vs Centre Back — Half-Space Threat. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Forest vs Arsenal

Premier League · 20261018 · 3-4-2-1 vs 4-2-3-1

Predicted 1-1Winner: DrawWinning H 2-1 (6.1%)Winning A 1-2 (9.8%)Confidence 0.75λ home/away 1.12 / 1.79Most-likely P 11.0%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Forest vs Arsenal. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 3-4-2-1 vs 4-2-3-1 (duel lambdas: Home 1.150, Away 1.742). Final calibrated Poisson lambdas: Home 1.117, Away 1.792 (most likely score 1-1).

Home Starters 11/11 XGBoost (9 FBref) Away Starters 11/11 XGBoost (11 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.15 / 1.74 Final λ (H/A) 1.12 / 1.79

FBref Team Priors: Nottingham Forest (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 1.379) vs Arsenal (Gls/90: 2.0, GA/90: 2.0 → E[λ]: 1.379) | Blend Weight: 0.192

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Forest	Arsenal
League position	15.00	2.00
Points per game	1.09	1.92
Attack index	0.85	1.04
Defense (concede) index	0.85	0.97
Form multiplier	0.96	1.04
Standing λ multiplier	0.88	1.01

Lambda decomposition (how the goal rates were built)

Stage	Forest λ	Arsenal λ
FBref Squad Prior E[λ] (weight 0.192)	1.379	1.379
Pre-competitive-floor	1.117	1.792
Pre-calibration (Micro-duels)	1.150	1.742
Final λ	1.117	1.792

competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Falsecompetitive underdog floor applied Falsestage discount 1.0

Scoreline distribution & markets

Top 5 scorelines	P
1-1	11.0%
1-2	9.8%
0-1	9.7%
0-2	8.8%
2-1	6.1%

Home win	23.5%
Draw	23.3%
Away win	53.1%
Over 2.5	55.6%
BTTS yes	56.1%

Pregame fair line — favorite Arsenal (53.1%), fair total 2.91, fair spread -0.68

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Forest — projected line vs season baseline (3-4-2-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Liam Delap F 79 0 0.27 0.18 51% 2 1.2 6.8 51.6 29 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0102, quality 0.764)	`market_value_eur=37000000` `quality_factor=0.7641` `boost_factor=1.0102` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0017)	`blend_weight=0.0017` `history_matches=33` `season_overlap_discount=1.0` `per90_to_match_factor=0.8746` `prior_reliability=0.02` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Bruno Guimarães, Riccardo Calafiori', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle', 'duel_win_rate': 0.219, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.9244, 'skill_edge': -0.0554, 'height_edge': 0.004, 'attacker_skill': 0.7641, 'defender_skill': 0.8873, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Bruno Guimarães', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}, {'player': 'Riccardo Calafiori', 'market_value': 63000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8852}], 'xg_override': 0.3386}	totalShots: 2.1849 -> 2.0197 shotOffTarget: 1.2776 -> 1.181 onTargetScoringAttempt: 0.9041 -> 0.8358 bigChanceMissed: 0.6184 -> 0.5716 expectedGoals: 0.3663 -> 0.3386 goals: 0.3324 -> 0.3073 expectedGoalsOnTarget: 0.1576 -> 0.1457
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.3386 -> 0.2663 expected_assists: 0.0611 -> 0.0496
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 77.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.2338` `goal_rate=0.3073` `scorer_weight=0.2449`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0484` `assist_rate=0.0` `provider_weight=0.0397`	-

Morgan Gibbs-White M 89 0 0.18 0.27 34% 2 1.8 7.1 61.2 35 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0117, quality 0.878)	`market_value_eur=76000000` `quality_factor=0.8781` `boost_factor=1.0117` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0090)	`blend_weight=0.009` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.9884` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 3.8286 -> 3.8119
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Bruno Guimarães', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.404, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9328, 'skill_edge': -0.0017, 'height_edge': -0.044, 'attacker_skill': 0.8855, 'defender_skill': 0.8893, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Bruno Guimarães', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}], 'xg_override': 0.2277}	aerialWon: 0.5532 -> 0.8 totalShots: 2.1674 -> 2.0218 shotOffTarget: 1.3318 -> 1.2423 onTargetScoringAttempt: 0.8342 -> 0.7781 bigChanceMissed: 0.3484 -> 0.325 goals: 0.2755 -> 0.257 expectedGoalsOnTarget: 0.2542 -> 0.2371 expectedGoals: 0.2441 -> 0.2277
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.2277 -> 0.179 expected_assists: 0.1898 -> 0.1542
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 83.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1639` `goal_rate=0.257` `scorer_weight=0.1979`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1429` `assist_rate=0.1812` `provider_weight=0.1487`	-

Dan Ndoye F 68 1 0.23 0.07 219% 2 0.7 6.9 30.6 31 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0103, quality 0.746)	`market_value_eur=34000000` `quality_factor=0.7455` `boost_factor=1.0103` `reference_quality=0.505` `history_matches=30` `history_shrink=0.1429`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0087)	`blend_weight=0.0087` `history_matches=30` `season_overlap_discount=1.0` `per90_to_match_factor=0.7557` `prior_reliability=0.0933` `competition=Premier League`	totalCross: 1.1036 -> 1.094
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Bruno Guimarães, Riccardo Calafiori', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle', 'duel_win_rate': 0.202, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.9004, 'skill_edge': -0.0638, 'height_edge': -0.004, 'attacker_skill': 0.7455, 'defender_skill': 0.8873, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Bruno Guimarães', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}, {'player': 'Riccardo Calafiori', 'market_value': 63000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8852}], 'xg_override': 0.2886}	totalShots: 2.1499 -> 1.9358 shotOffTarget: 1.2583 -> 1.133 onTargetScoringAttempt: 0.8916 -> 0.8028 bigChanceMissed: 0.8156 -> 0.7344 goals: 0.3904 -> 0.3515 expectedGoals: 0.3205 -> 0.2886 expectedGoalsOnTarget: 0.2926 -> 0.2635
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.2886 -> 0.2269 expected_assists: 0.0904 -> 0.0734
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 77.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.203` `goal_rate=0.3515` `scorer_weight=0.2691`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0708` `assist_rate=0.0084` `provider_weight=0.0587`	-

Ibrahim Sangaré M 81 0 0.03 0.02 48% 1 0.4 6.9 53.9 24 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0055, quality 0.641)	`market_value_eur=21000000` `quality_factor=0.6408` `boost_factor=1.0055` `reference_quality=0.505` `history_matches=32` `history_shrink=0.1351`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0059)	`blend_weight=0.0059` `history_matches=32` `season_overlap_discount=1.0` `per90_to_match_factor=0.9008` `prior_reliability=0.0667` `competition=Premier League`	interceptionWon: 1.0641 -> 1.0579
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ezri Konsa', 'matchup_zone': 'Left Wingback vs Right Back — Overlapping Run', 'duel_win_rate': 0.36, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.941, 'skill_edge': -0.0721, 'height_edge': 0.032, 'attacker_skill': 0.7176, 'defender_skill': 0.8778, 'attacker_role': 'LWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Ezri Konsa', 'market_value': 53000000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.8778}], 'xg_override': 0.0368}	aerialWon: 1.1402 -> 1.3 totalShots: 0.5591 -> 0.5261 shotOffTarget: 0.401 -> 0.3773 onTargetScoringAttempt: 0.1552 -> 0.146
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0368 -> 0.0289 expected_assists: 0.0481 -> 0.0391
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 78.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0285` `goal_rate=0.0444` `scorer_weight=0.0314`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0383` `assist_rate=0.047` `provider_weight=0.0387`	-

James McAtee M 84 0 0.05 0.02 182% 1 0.4 6.9 26.9 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0074, quality 0.629)	`market_value_eur=19800000` `quality_factor=0.6291` `boost_factor=1.0074` `reference_quality=0.505` `history_matches=20` `history_shrink=0.2`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0156)	`blend_weight=0.0156` `history_matches=20` `season_overlap_discount=1.0` `per90_to_match_factor=0.938` `prior_reliability=0.12` `competition=Premier League`	totalShots: 1.2371 -> 1.2178 fouls: 1.027 -> 1.011 interceptionWon: 0.9298 -> 0.9153 wasFouled: 1.3298 -> 1.3416 totalCross: 3.2691 -> 3.275 onTargetScoringAttempt: 0.3513 -> 0.3458
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Gabriel Magalhães', 'matchup_zone': 'Right Wingback vs Left Back — Overlapping Run', 'duel_win_rate': 0.275, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8163, 'skill_edge': -0.085, 'height_edge': -0.04, 'attacker_skill': 0.7044, 'defender_skill': 0.8933, 'attacker_role': 'RWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Gabriel Magalhães', 'market_value': 70000000, 'role': 'LB', 'price_index': 0.7, 'adjusted_quality': 0.8933}], 'xg_override': 0.069}	aerialWon: 0.6209 -> 1.1 totalShots: 1.2315 -> 1.0053 shotOffTarget: 0.8857 -> 0.723 onTargetScoringAttempt: 0.3458 -> 0.2823 goals: 0.0922 -> 0.0753 expectedGoals: 0.0845 -> 0.069
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.069 -> 0.0542 expected_assists: 0.133 -> 0.108
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 73.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0528` `goal_rate=0.0753` `scorer_weight=0.0571`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1024` `assist_rate=0.0981` `provider_weight=0.0874`	-

Daniel Muñoz M 78 0 0.10 0.06 67% 1 0.8 6.8 51.1 28 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0059, quality 0.660)	`market_value_eur=23000000` `quality_factor=0.6595` `boost_factor=1.0059` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0033)	`blend_weight=0.0033` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.8721` `prior_reliability=0.04` `competition=Premier League`	totalCross: 1.7313 -> 1.7255
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Bruno Guimarães', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.361, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8688, 'skill_edge': -0.0933, 'height_edge': 0.004, 'attacker_skill': 0.6821, 'defender_skill': 0.8893, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Bruno Guimarães', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}], 'xg_override': 0.1303}	aerialWon: 2.163 -> 1.1 totalShots: 1.4002 -> 1.2165 shotOffTarget: 0.9352 -> 0.8125 bigChanceMissed: 0.4702 -> 0.4085 onTargetScoringAttempt: 0.4632 -> 0.4024 goals: 0.1529 -> 0.1328 expectedGoals: 0.15 -> 0.1303 expectedGoalsOnTarget: 0.1183 -> 0.1028
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1303 -> 0.1024 expected_assists: 0.0858 -> 0.0697
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 82.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0973` `goal_rate=0.1328` `scorer_weight=0.1038`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0673` `assist_rate=0.0844` `provider_weight=0.069`	-

Murillo D 87 0 0.07 0.05 59% 1 0.6 6.9 54.2 23 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0131, quality 0.821)	`market_value_eur=49000000` `quality_factor=0.8206` `boost_factor=1.0131` `reference_quality=0.505` `history_matches=31` `history_shrink=0.1389`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0120)	`blend_weight=0.012` `history_matches=31` `season_overlap_discount=1.0` `per90_to_match_factor=0.9684` `prior_reliability=0.1333` `competition=Premier League`	fouls: 1.4394 -> 1.4279 totalShots: 1.2877 -> 1.278 interceptionWon: 1.2225 -> 1.2136 wonTackle: 1.1315 -> 1.1237
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Martin Ødegaard', 'matchup_zone': 'LW vs CB'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0952 -> 0.0748 expected_assists: 0.0361 -> 0.0293
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 82.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0721` `goal_rate=0.0049` `scorer_weight=0.0598`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0289` `assist_rate=0.1092` `provider_weight=0.0366`	-

Ousmane Diomande D 87 0 0.06 0.02 196% 1 0.3 6.8 17.8 11 FBref Prior D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0492, quality 0.800)	`market_value_eur=44000000` `quality_factor=0.8002` `boost_factor=1.0492` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kai Havertz', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0789 -> 0.062
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 81.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0601` `goal_rate=0.0` `scorer_weight=0.0496`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0067` `assist_rate=0.0` `provider_weight=0.0058`	-

Jair D 83 0 0.02 0.02 ok 0 0.2 6.8 38.2 21 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0151, quality 0.686)	`market_value_eur=26000000` `quality_factor=0.6858` `boost_factor=1.0151` `reference_quality=0.505` `history_matches=13` `history_shrink=0.2778`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Bukayo Saka', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 14 by projected rating, fifa_overall 80.0)	`match_squad_rank=14` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0154` `goal_rate=0.0` `scorer_weight=0.0124`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0045` `assist_rate=0.0` `provider_weight=0.004`	-

Matz Sels G 90 0 0.00 - n/a 0.0 7.0 56.4 27 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9979, quality 0.446)	`market_value_eur=3700000` `quality_factor=0.4461` `boost_factor=0.9979` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0103)	`blend_weight=0.0103` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.9956` `prior_reliability=0.1333` `competition=Premier League`	saves: 3.0068 -> 2.9963
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kai Havertz, Bukayo Saka', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.7419}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	shots_faced: 16.5 -> 17.0 expected_saves: 4.03 -> 4.16 high_pressure_saves: 1.39 -> 1.43
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 81.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.1226`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.6637`	-

Neco Williams D 89 0 0.07 0.05 40% 1 1.2 6.8 58.0 35 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0064, quality 0.710)	`market_value_eur=29000000` `quality_factor=0.71` `boost_factor=1.0064` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0090)	`blend_weight=0.009` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.9912` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 1.374 -> 1.3929 totalCross: 5.7284 -> 5.7125 interceptionWon: 0.953 -> 0.9444 totalShots: 1.1762 -> 1.1701
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kai Havertz', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.12 (squad xG sum 1.4209, xG scale 0.7861, xA scale 0.8123, team xA 0.73 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.117` `squad_xg_sum=1.4209` `squad_xa_sum=0.8939` `team_expected_assists=0.726` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0861 -> 0.0677 expected_assists: 0.1518 -> 0.1233
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 80.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0655` `goal_rate=0.0291` `scorer_weight=0.0542`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.116` `assist_rate=0.0438` `provider_weight=0.0986`	-

Team totals 1 1.12 12 6.9

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
John Victor	G	4.5m	11.3%	Market-value substitute prior
Nikola Milenković	D	14.6m	23.7%	Market-value substitute prior
Ola Aina	D	14.2m	23.0%	Market-value substitute prior
Omari Hutchinson	M	23.0m	54.6%	Market-value substitute prior
Xaver Schlager	M	19.5m	51.0%	Market-value substitute prior
Callum Hudson Odoi	F	26.5m	61.2%	Market-value substitute prior

Arsenal — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Bukayo Saka F 84 0 0.36 0.30 ok 2 1.9 7.3 59.4 32 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0140, quality 0.897)	`market_value_eur=115000000` `quality_factor=0.8968` `boost_factor=1.014` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.9323` `prior_reliability=0.1133` `competition=Premier League`	totalCross: 3.3589 -> 3.3727 wasFouled: 2.495 -> 2.5068 fouls: 1.0433 -> 1.0342 goals: 0.353 -> 0.3595 wonTackle: 0.9672 -> 0.9731
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Murillo, Ousmane Diomande', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.184, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.9768, 'skill_edge': 0.0182, 'height_edge': -0.034, 'attacker_skill': 0.8968, 'defender_skill': 0.8564, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Murillo', 'market_value': 49000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8633}, {'player': 'Ousmane Diomande', 'market_value': 44000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8495}], 'xg_override': 0.355}	totalShots: 2.4963 -> 2.4384 shotOffTarget: 1.4417 -> 1.4083 onTargetScoringAttempt: 1.0532 -> 1.0288 bigChanceMissed: 0.5891 -> 0.5754 expectedGoalsOnTarget: 0.5069 -> 0.4951 expectedGoals: 0.3634 -> 0.355 goals: 0.3595 -> 0.3512
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 87.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.2992` `goal_rate=0.3512` `scorer_weight=0.3522`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1838` `assist_rate=0.1132` `provider_weight=0.1625`	-

Bruno Guimarães M 81 0 0.12 0.14 ok 2 0.9 7.2 53.5 21 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0148, quality 0.879)	`market_value_eur=77000000` `quality_factor=0.8792` `boost_factor=1.0148` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0011)	`blend_weight=0.0011` `history_matches=33` `season_overlap_discount=1.0` `per90_to_match_factor=0.9046` `prior_reliability=0.0133` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Liam Delap, Dan Ndoye, Morgan Gibbs-White', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 86.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1096` `goal_rate=0.1262` `scorer_weight=0.1233`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.139` `assist_rate=0.0981` `provider_weight=0.1198`	-

Declan Rice M 86 0 0.04 0.04 ok 1 0.6 7.3 59.7 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0125, quality 0.897)	`market_value_eur=114000000` `quality_factor=0.8967` `boost_factor=1.0125` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0078)	`blend_weight=0.0078` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.9541` `prior_reliability=0.1133` `competition=Premier League`	totalCross: 3.4043 -> 3.4128
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ibrahim Sangaré', 'matchup_zone': 'Midfield lane'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 88.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.044` `goal_rate=0.0506` `scorer_weight=0.0477`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1333` `assist_rate=0.1613` `provider_weight=0.1574`	-

Riccardo Calafiori D 79 0 0.18 0.09 102% 1 0.8 6.8 44.7 23 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0143, quality 0.859)	`market_value_eur=63000000` `quality_factor=0.8589` `boost_factor=1.0143` `reference_quality=0.505` `history_matches=32` `history_shrink=0.1351`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0111)	`blend_weight=0.0111` `history_matches=32` `season_overlap_discount=1.0` `per90_to_match_factor=0.8804` `prior_reliability=0.1267` `competition=Premier League`	goalAssist: 0.153 -> 0.1616 wonTackle: 0.963 -> 0.9574
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Liam Delap', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 82.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1674` `goal_rate=0.154` `scorer_weight=0.1605`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0768` `assist_rate=0.1616` `provider_weight=0.0999`	-

Gabriel Magalhães D 90 0 0.10 0.07 30% 1 0.7 6.9 59.6 30 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0128, quality 0.871)	`market_value_eur=70000000` `quality_factor=0.8707` `boost_factor=1.0128` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0101)	`blend_weight=0.0101` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=0.9996` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 0.7434 -> 0.7359 fouls: 0.9256 -> 0.9314 interceptionWon: 0.9588 -> 0.9642
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'James McAtee', 'matchup_zone': 'Right Wingback vs Left Back — Overlapping Run'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 88.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0907` `goal_rate=0.0078` `scorer_weight=0.0761`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0431` `assist_rate=0.0305` `provider_weight=0.0378`	-

Ezri Konsa D 90 0 0.05 0.04 ok 0 0.3 6.8 52.4 28 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0112, quality 0.834)	`market_value_eur=53000000` `quality_factor=0.834` `boost_factor=1.0112` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0005)	`blend_weight=0.0005` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9999` `prior_reliability=0.0067` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ibrahim Sangaré', 'matchup_zone': 'Left Wingback vs Right Back — Overlapping Run'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 84.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0498` `goal_rate=0.0` `scorer_weight=0.0409`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0095` `assist_rate=0.0` `provider_weight=0.008`	-

Ben White D 73 0 0.03 0.01 170% 0 0.1 6.7 34.6 24 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0155, quality 0.732)	`market_value_eur=32000000` `quality_factor=0.732` `boost_factor=1.0155` `reference_quality=0.505` `history_matches=17` `history_shrink=0.2273`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0197)	`blend_weight=0.0197` `history_matches=17` `season_overlap_discount=1.0` `per90_to_match_factor=0.8161` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 1.8276 -> 1.8398 wasFouled: 0.8817 -> 0.8724 wonTackle: 1.081 -> 1.0758
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Liam Delap', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 14 by projected rating, fifa_overall 86.0)	`match_squad_rank=14` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.029` `goal_rate=0.0` `scorer_weight=0.0235`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0794` `assist_rate=0.285` `provider_weight=0.1034`	-

David Raya G 90 0 0.00 - n/a 0.0 6.9 57.4 29 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0073, quality 0.739)	`market_value_eur=33000000` `quality_factor=0.7389` `boost_factor=1.0073` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0090)	`blend_weight=0.009` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	saves: 2.9583 -> 2.9361 wasFouled: 0.6282 -> 0.6225
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Liam Delap, Dan Ndoye', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.1498}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	shots_faced: 10.9 -> 10.6 expected_saves: 2.67 -> 2.59 high_pressure_saves: 0.92 -> 0.89
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 88.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.7121`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=1.1173`	-

Martin Ødegaard M 80 0 0.19 0.10 94% 2 0.6 7.2 52.1 23 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0164, quality 0.888)	`market_value_eur=88000000` `quality_factor=0.8879` `boost_factor=1.0164` `reference_quality=0.505` `history_matches=30` `history_shrink=0.1429`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0099)	`blend_weight=0.0099` `history_matches=30` `season_overlap_discount=1.0` `per90_to_match_factor=0.8869` `prior_reliability=0.1067` `competition=Premier League`	wonTackle: 0.5436 -> 0.5547 totalShots: 1.4303 -> 1.4379 totalCross: 1.1395 -> 1.1447
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Murillo', 'matchup_zone': 'LW vs CB', 'duel_win_rate': 0.411, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9869, 'skill_edge': 0.0111, 'height_edge': -0.02, 'attacker_skill': 0.8879, 'defender_skill': 0.8633, 'attacker_role': 'LW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Murillo', 'market_value': 49000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8633}], 'xg_override': 0.1944}`	aerialWon: 0.0701 -> 1.0 totalShots: 1.4379 -> 1.4191 shotOffTarget: 0.8154 -> 0.8047 onTargetScoringAttempt: 0.6143 -> 0.6063
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 86.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1769` `goal_rate=0.1812` `scorer_weight=0.1842`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1623` `assist_rate=0.1685` `provider_weight=0.1713`	-

Eberechi Eze M 74 0 0.15 0.09 71% 2 1.1 6.8 49.9 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0126, quality 0.859)	`market_value_eur=63000000` `quality_factor=0.8589` `boost_factor=1.0126` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0036)	`blend_weight=0.0036` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.8207` `prior_reliability=0.0467` `competition=Premier League`	wasFouled: 1.8442 -> 1.8502
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Morgan Gibbs-White', 'matchup_zone': 'CAM vs CM', 'duel_win_rate': 0.439, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 1.0275, 'skill_edge': -0.0093, 'height_edge': 0.028, 'attacker_skill': 0.865, 'defender_skill': 0.8855, 'attacker_role': 'CAM', 'attacker_mv_price_index': 0.95, 'defender_adjusted_parts': [{'player': 'Morgan Gibbs-White', 'market_value': 76000000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.8855}], 'xg_override': 0.1479}	aerialWon: 0.3706 -> 1.0 totalShots: 1.4618 -> 1.502 shotOffTarget: 0.9413 -> 0.9672 onTargetScoringAttempt: 0.5194 -> 0.5337 expectedGoalsOnTarget: 0.3316 -> 0.3407
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 84.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1377` `goal_rate=0.1415` `scorer_weight=0.1433`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0637` `assist_rate=0.0608` `provider_weight=0.0632`	-

Kai Havertz F 77 1 0.40 0.22 79% 3 1.3 6.9 50.2 23 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0224, quality 0.848)	`market_value_eur=58000000` `quality_factor=0.8478` `boost_factor=1.0224` `reference_quality=0.505` `history_matches=18` `history_shrink=0.2174`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0188)	`blend_weight=0.0188` `history_matches=18` `season_overlap_discount=1.0` `per90_to_match_factor=0.8524` `prior_reliability=0.1333` `competition=Premier League`	totalOffside: 1.2009 -> 1.2345 totalShots: 2.4483 -> 2.4182 wasFouled: 1.0734 -> 1.0853 totalCross: 0.4955 -> 0.4862 fouls: 0.9056 -> 0.9126
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Murillo, Ousmane Diomande', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.222, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 1.0326, 'skill_edge': -0.0038, 'height_edge': 0.026, 'attacker_skill': 0.8478, 'defender_skill': 0.8564, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Murillo', 'market_value': 49000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8633}, {'player': 'Ousmane Diomande', 'market_value': 44000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8495}], 'xg_override': 0.3991}	totalShots: 2.4457 -> 2.5255 shotOffTarget: 1.4561 -> 1.5036 onTargetScoringAttempt: 0.9896 -> 1.0219 bigChanceMissed: 0.7184 -> 0.7418 expectedGoalsOnTarget: 0.4299 -> 0.4439 expectedGoals: 0.3865 -> 0.3991 goals: 0.3726 -> 0.3847
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.79 (squad xG sum 1.7892, xG scale 1.0015, xA scale 1.0069, team xA 1.16 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.792` `squad_xg_sum=1.7892` `squad_xa_sum=1.1568` `team_expected_assists=1.1648` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 81.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.3295` `goal_rate=0.3847` `scorer_weight=0.387`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.075` `assist_rate=0.0836` `provider_weight=0.0817`	-

Team totals 1 1.79 16 7.0

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Illan Meslier	G	4.6m	11.9%	Market-value substitute prior
William Saliba	D	15.4m	38.0%	Market-value substitute prior
Jurriën Timber	D	15.3m	37.7%	Market-value substitute prior
Martín Zubimendi	M	18.9m	72.8%	Market-value substitute prior
Max Dowman	M	23.5m	57.0%	Market-value substitute prior
Viktor Gyökeres	F	26.9m	63.4%	Market-value substitute prior

Antagonist / tactical summary

Midfield is evenly contested — Morgan Gibbs-White vs Declan Rice will be the pivotal duel. Total antagonist xG — Home: 1.09, Away: 1.10. Home danger zone: Shadow Striker vs CDM + CB — Zone 14 Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Midfield — possession 48.2% / 51.8%, dominance ratio 0.45, key duel: Morgan Gibbs-White vs Declan Rice

Liam Delap xG 0.34Dan Ndoye xG 0.29Ibrahim Sangaré xG 0.04Martin Ødegaard xG 0.19Eberechi Eze xG 0.15Kai Havertz xG 0.40

Summary

Antagonist matchup analysis for Forest vs Arsenal. Most likely scoreline: 1-1. Lambdas derived from 6 home and 4 away individual player duels. Head-to-head: Forest 2W-2D-6L vs Arsenal over 10 meetings.

Tactical parity between 3-4-2-1 and 4-2-3-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Morgan Gibbs-White vs Declan Rice will be the pivotal duel. Total antagonist xG — Home: 1.09, Away: 1.10. Home danger zone: Shadow Striker vs CDM + CB — Zone 14 Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Brighton vs Crystal Palace

Premier League · 20261018 · 4-2-3-1 vs 3-4-2-1

Predicted 1-1Winner: DrawWinning H 1-0 (10.0%)Winning A 0-1 (8.5%)Confidence 0.7λ home/away 1.43 / 1.22Most-likely P 12.4%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Brighton vs Crystal Palace. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-2-3-1 vs 3-4-2-1 (duel lambdas: Home 1.440, Away 1.207). Final calibrated Poisson lambdas: Home 1.430, Away 1.216 (most likely score 1-1).

Home Starters 11/11 XGBoost (9 FBref) Away Starters 11/11 XGBoost (8 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.44 / 1.21 Final λ (H/A) 1.43 / 1.22

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Brighton	Crystal Palace
League position	3.00	12.00
Points per game	1.75	1.25
Attack index	1.34	1.01
Defense (concede) index	1.12	0.85
Form multiplier	1.03	0.97
Standing λ multiplier	1.19	0.94

Lambda decomposition (how the goal rates were built)

Stage	Brighton λ	Crystal Palace λ
Pre-competitive-floor	1.430	1.216
Pre-calibration (Micro-duels)	1.440	1.207
Final λ	1.430	1.216

Scoreline distribution & markets

Top 5 scorelines	P
1-1	12.4%
1-0	10.0%
2-1	8.8%
0-1	8.5%
1-2	7.5%

Home win	41.9%
Draw	26.2%
Away win	31.9%
Over 2.5	49.3%
BTTS yes	53.6%

Pregame fair line — favorite Brighton (41.9%), fair total 2.65, fair spread 0.21

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Brighton — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Charalampos Kostoulas F 77 1 0.28 0.10 165% 2 0.9 6.8 42.7 27 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0136, quality 0.796)	`market_value_eur=43000000` `quality_factor=0.7957` `boost_factor=1.0136` `reference_quality=0.505` `history_matches=27` `history_shrink=0.1562`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0081)	`blend_weight=0.0081` `history_matches=27` `season_overlap_discount=1.0` `per90_to_match_factor=0.8567` `prior_reliability=0.08` `competition=Premier League`	wasFouled: 1.1189 -> 1.1098 wonTackle: 0.7157 -> 0.7099
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Jaydee Canvot, Chris Richards', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.201, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 1.0009, 'skill_edge': 0.0126, 'height_edge': -0.012, 'attacker_skill': 0.7957, 'defender_skill': 0.7676, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Jaydee Canvot', 'market_value': 32000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.7941}, {'player': 'Chris Richards', 'market_value': 25000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.7411}], 'xg_override': 0.2502}	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.2502 -> 0.2767 expected_assists: 0.063 -> 0.0697
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 12 by projected rating, fifa_overall 73.0)	`match_squad_rank=12` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.2417` `goal_rate=0.3154` `scorer_weight=0.3368`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0673` `assist_rate=0.2724` `provider_weight=0.0871`	-

Maxim De Cuyper D 79 0 0.19 0.11 65% 1 0.6 6.8 50.2 35 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0075, quality 0.710)	`market_value_eur=29000000` `quality_factor=0.71` `boost_factor=1.0075` `reference_quality=0.505` `history_matches=36` `history_shrink=0.122`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0100)	`blend_weight=0.01` `history_matches=36` `season_overlap_discount=1.0` `per90_to_match_factor=0.8724` `prior_reliability=0.1267` `competition=Premier League`	totalCross: 4.354 -> 4.3702 totalShots: 0.8201 -> 0.8354 fouls: 0.907 -> 0.9163 wasFouled: 0.743 -> 0.7355
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Quinten Timber', 'matchup_zone': 'Right Wingback vs Left Back — Overlapping Run'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.1696 -> 0.1877 expected_assists: 0.1754 -> 0.194
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 78.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1711` `goal_rate=0.2392` `scorer_weight=0.2346`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1763` `assist_rate=0.2884` `provider_weight=0.2425`	-

Yasin Ayari M 86 0 0.17 0.05 224% 1 0.8 6.8 48.7 24 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0108, quality 0.786)	`market_value_eur=41000000` `quality_factor=0.7859` `boost_factor=1.0108` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0039)	`blend_weight=0.0039` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.9607` `prior_reliability=0.0467` `competition=Premier League`	fouls: 0.5778 -> 0.5862
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jørgen Strand Larsen, Yéremy Pino, Adam Wharton', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.1494 -> 0.1653 expected_assists: 0.0667 -> 0.0738
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 79.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1524` `goal_rate=0.1161` `scorer_weight=0.1377`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0711` `assist_rate=0.055` `provider_weight=0.0664`	-

Diego Gómez M 81 0 0.20 0.14 48% 2 1.4 7.0 54.8 26 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0079, quality 0.732)	`market_value_eur=32000000` `quality_factor=0.732` `boost_factor=1.0079` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0096)	`blend_weight=0.0096` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=0.9051` `prior_reliability=0.1267` `competition=Premier League`	totalShots: 1.912 -> 1.9293 onTargetScoringAttempt: 0.6323 -> 0.6396
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Daichi Kamada', 'matchup_zone': 'Midfield lane'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.1826 -> 0.202 expected_assists: 0.0884 -> 0.0978
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 78.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1829` `goal_rate=0.2178` `scorer_weight=0.2326`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0932` `assist_rate=0.1116` `provider_weight=0.1142`	-

Chema Andrés M 87 0 0.10 0.06 71% 1 0.3 6.8 23.6 9 XGBoost M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0290, quality 0.660)	`market_value_eur=23000000` `quality_factor=0.6595` `boost_factor=1.029` `reference_quality=0.505` `history_matches=3` `history_shrink=0.625`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Adam Wharton', 'matchup_zone': 'CAM vs CM', 'duel_win_rate': 0.366, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9209, 'skill_edge': -0.0978, 'height_edge': 0.044, 'attacker_skill': 0.6704, 'defender_skill': 0.8877, 'attacker_role': 'CAM', 'attacker_mv_price_index': 0.95, 'defender_adjusted_parts': [{'player': 'Adam Wharton', 'market_value': 79000000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.8877}], 'xg_override': 0.0891}	aerialWon: 1.989 -> 1.4 totalShots: 1.0938 -> 1.0073 shotOffTarget: 0.7775 -> 0.716 onTargetScoringAttempt: 0.3163 -> 0.2913 expectedGoalsOnTarget: 0.316 -> 0.291 expectedGoals: 0.0967 -> 0.0891 goals: 0.0885 -> 0.0815
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.0891 -> 0.0986 expected_assists: 0.0559 -> 0.0619
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 86.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0939` `goal_rate=0.0815` `scorer_weight=0.0932`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.06` `assist_rate=0.0696` `provider_weight=0.0699`	-

Pascal Groß M 90 0 0.15 0.12 30% 1 1.0 7.3 59.7 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9966, quality 0.437)	`market_value_eur=2300000` `quality_factor=0.4374` `boost_factor=0.9966` `reference_quality=0.505` `history_matches=25` `history_shrink=0.1667`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0144)	`blend_weight=0.0144` `history_matches=25` `season_overlap_discount=1.0` `per90_to_match_factor=0.9964` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 6.0142 -> 6.0569 totalShots: 1.5262 -> 1.5114 interceptionWon: 0.6354 -> 0.6262
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Jaydee Canvot', 'matchup_zone': 'LW vs CB', 'duel_win_rate': 0.231, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.7229, 'skill_edge': -0.1605, 'height_edge': -0.028, 'attacker_skill': 0.4374, 'defender_skill': 0.7941, 'attacker_role': 'LW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Jaydee Canvot', 'market_value': 32000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.7941}], 'xg_override': 0.1364}	totalShots: 1.5251 -> 1.1025 aerialWon: 0.7374 -> 1.1 shotOffTarget: 0.9601 -> 0.6941 onTargetScoringAttempt: 0.565 -> 0.4084 goals: 0.2328 -> 0.1683 expectedGoals: 0.1887 -> 0.1364 expectedGoalsOnTarget: 0.1477 -> 0.1068 bigChanceMissed: 0.0982 -> 0.071
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.1364 -> 0.1509 expected_assists: 0.1747 -> 0.1933
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 79.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1401` `goal_rate=0.1683` `scorer_weight=0.1767`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1758` `assist_rate=0.1736` `provider_weight=0.1924`	-

Olivier Boscagli D 85 0 0.01 0.01 54% 1 0.2 7.0 36.3 31 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0072, quality 0.616)	`market_value_eur=18500000` `quality_factor=0.6161` `boost_factor=1.0072` `reference_quality=0.505` `history_matches=18` `history_shrink=0.2174`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0188)	`blend_weight=0.0188` `history_matches=18` `season_overlap_discount=1.0` `per90_to_match_factor=0.9417` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 2.9674 -> 2.9558 fouls: 1.2779 -> 1.2893 onTargetScoringAttempt: 0.0864 -> 0.0936 wonTackle: 0.8308 -> 0.824 interceptionWon: 1.1323 -> 1.1376
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Tyrick Mitchell', 'matchup_zone': 'Left Wingback vs Right Back — Overlapping Run'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_assists: 0.0756 -> 0.0837
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 78.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0079` `goal_rate=0.0` `scorer_weight=0.0069`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0803` `assist_rate=0.0232` `provider_weight=0.067`	-

Lewis Dunk D 90 0 0.05 0.04 ok 1 0.6 7.1 53.4 27 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9979, quality 0.445)	`market_value_eur=3600000` `quality_factor=0.4449` `boost_factor=0.9979` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0101)	`blend_weight=0.0101` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	fouls: 0.6454 -> 0.654 totalCross: 0.2445 -> 0.2521 totalShots: 0.5563 -> 0.5507
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Yéremy Pino', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.0486 -> 0.0538
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 77.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0524` `goal_rate=0.2108` `scorer_weight=0.0673`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0223` `assist_rate=0.0` `provider_weight=0.0181`	-

Luka Vušković D 88 0 0.15 0.07 128% 1 0.9 6.9 41.0 11 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0497, quality 0.869)	`market_value_eur=69000000` `quality_factor=0.8692` `boost_factor=1.0497` `reference_quality=0.505` `history_matches=6` `history_shrink=0.4545`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0394)	`blend_weight=0.0394` `history_matches=6` `season_overlap_discount=1.0` `per90_to_match_factor=0.9826` `prior_reliability=0.1333` `competition=Premier League`	wasFouled: 1.3905 -> 1.3357 interceptionWon: 1.1102 -> 1.0665 totalShots: 1.0579 -> 1.0936 wonTackle: 0.8248 -> 0.8117 totalCross: 0.3105 -> 0.2983 fouls: 1.2589 -> 1.248 goals: 0.1553 -> 0.1492 onTargetScoringAttempt: 0.1553 -> 0.1492
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jørgen Strand Larsen', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_goals: 0.1358 -> 0.1503
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 81.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1396` `goal_rate=0.1492` `scorer_weight=0.1611`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0125` `assist_rate=0.0` `provider_weight=0.0103`	-

Ferdi Kadıoğlu D 87 0 0.04 0.05 ok 1 0.7 6.8 56.7 29 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0090, quality 0.786)	`market_value_eur=41000000` `quality_factor=0.7859` `boost_factor=1.009` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jørgen Strand Larsen', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	expected_assists: 0.0705 -> 0.078
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 80.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0416` `goal_rate=0.0355` `scorer_weight=0.0411`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.075` `assist_rate=0.0` `provider_weight=0.0624`	-

Bart Verbruggen G 90 0 0.00 - n/a 0.0 7.0 58.7 30 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0079, quality 0.764)	`market_value_eur=37000000` `quality_factor=0.7641` `boost_factor=1.0079` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	saves: 2.9474 -> 2.9258
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jørgen Strand Larsen, Yéremy Pino', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.207}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.43 (squad xG sum 1.2924, xG scale 1.1065, xA scale 1.1065, team xA 0.97 = 0.68 x lambda)	`scoreline_reset=true` `team_lambda=1.43` `squad_xg_sum=1.2924` `squad_xa_sum=0.8726` `team_expected_assists=0.9656` `team_xa_to_xg_ratio=0.6752`	shots_faced: 11.5 -> 11.6 expected_saves: 2.82 -> 2.84
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 81.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.5787`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.9927`	-

Team totals 1 1.43 14 6.9

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Tom McGill	G	4.4m	10.9%	Market-value substitute prior
Mats Wieffer	D	14.5m	23.6%	Market-value substitute prior
Pascal Struijk	D	14.4m	33.9%	Market-value substitute prior
Jack Hinshelwood	M	23.2m	33.2%	Market-value substitute prior
Georginio Rutter	M	23.1m	33.0%	Market-value substitute prior
Yankuba Minteh	F	26.8m	62.9%	Market-value substitute prior

Crystal Palace — projected line vs season baseline (3-4-2-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Jørgen Strand Larsen F 75 1 0.33 0.14 141% 2 1.2 6.8 54.7 30 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0090, quality 0.786)	`market_value_eur=41000000` `quality_factor=0.7859` `boost_factor=1.009` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0046)	`blend_weight=0.0046` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.8336` `prior_reliability=0.0667` `competition=Premier League`	fouls: 0.9267 -> 0.934
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yasin Ayari, Ferdi Kadıoğlu', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle', 'duel_win_rate': 0.311, 'pressing_faced': 'HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 1.0602, 'skill_edge': -0.0191, 'height_edge': 0.06, 'attacker_skill': 0.7859, 'defender_skill': 0.8283, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Yasin Ayari', 'market_value': 41000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8177}, {'player': 'Ferdi Kadıoğlu', 'market_value': 41000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.839}], 'xg_override': 0.3663}	totalShots: 2.1619 -> 2.292 shotOffTarget: 1.3445 -> 1.4254 onTargetScoringAttempt: 0.8174 -> 0.8666 expectedGoals: 0.3455 -> 0.3663 goals: 0.2989 -> 0.3169 bigChanceMissed: 0.2739 -> 0.2904 expectedGoalsOnTarget: 0.2069 -> 0.2194
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.3663 -> 0.3292
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 12 by projected rating, fifa_overall 77.0)	`match_squad_rank=12` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.2805` `goal_rate=0.3169` `scorer_weight=0.2901`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0422` `assist_rate=0.0` `provider_weight=0.0345`	-

Yéremy Pino F 70 0 0.13 0.08 61% 1 0.9 7.0 53.3 32 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0060, quality 0.694)	`market_value_eur=27000000` `quality_factor=0.6941` `boost_factor=1.006` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0051)	`blend_weight=0.0051` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.7803` `prior_reliability=0.0733` `competition=Premier League`	totalCross: 2.7188 -> 2.7374 fouls: 0.8787 -> 0.885 wasFouled: 1.0177 -> 1.0233 totalOffside: 0.409 -> 0.4141
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yasin Ayari, Ferdi Kadıoğlu', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle', 'duel_win_rate': 0.206, 'pressing_faced': 'HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.9053, 'skill_edge': -0.0604, 'height_edge': -0.004, 'attacker_skill': 0.6941, 'defender_skill': 0.8283, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Yasin Ayari', 'market_value': 41000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8177}, {'player': 'Ferdi Kadıoğlu', 'market_value': 41000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.839}], 'xg_override': 0.1427}	totalShots: 1.492 -> 1.3507 shotOffTarget: 0.9455 -> 0.856 onTargetScoringAttempt: 0.5408 -> 0.4896 expectedGoalsOnTarget: 0.2655 -> 0.2404 goals: 0.2051 -> 0.1857 expectedGoals: 0.1576 -> 0.1427 bigChanceMissed: 0.1147 -> 0.1038
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.1427 -> 0.1282 expected_assists: 0.1331 -> 0.1196
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 79.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1203` `goal_rate=0.1857` `scorer_weight=0.1568`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1127` `assist_rate=0.0068` `provider_weight=0.0957`	-

Daichi Kamada M 84 0 0.07 0.07 ok 1 0.5 6.8 54.1 29 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0008, quality 0.526)	`market_value_eur=10300000` `quality_factor=0.526` `boost_factor=1.0008` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0111)	`blend_weight=0.0111` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.9324` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 1.4511 -> 1.4712 totalCross: 0.9645 -> 0.9745 totalShots: 1.0307 -> 1.0244
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yasin Ayari', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.373, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8864, 'skill_edge': -0.1253, 'height_edge': 0.048, 'attacker_skill': 0.5393, 'defender_skill': 0.8177, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Yasin Ayari', 'market_value': 41000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8177}], 'xg_override': 0.0766}	aerialWon: 0.2679 -> 1.1 totalShots: 1.0264 -> 0.9098 shotOffTarget: 0.652 -> 0.5779 onTargetScoringAttempt: 0.3744 -> 0.3319 bigChanceMissed: 0.1511 -> 0.1339 expectedGoalsOnTarget: 0.1258 -> 0.1115 goals: 0.0922 -> 0.0817 expectedGoals: 0.0864 -> 0.0766
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.0766 -> 0.0688 expected_assists: 0.1006 -> 0.0904
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 79.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0665` `goal_rate=0.0817` `scorer_weight=0.0716`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0864` `assist_rate=0.0959` `provider_weight=0.0876`	-

Tyrick Mitchell M 85 0 0.07 0.06 ok 1 0.6 6.9 57.8 31 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0047, quality 0.660)	`market_value_eur=23000000` `quality_factor=0.6595` `boost_factor=1.0047` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=0.9486` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 2.0587 -> 2.0489
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Olivier Boscagli', 'matchup_zone': 'Left Wingback vs Right Back — Overlapping Run', 'duel_win_rate': 0.394, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9909, 'skill_edge': 0.0218, 'height_edge': -0.028, 'attacker_skill': 0.7379, 'defender_skill': 0.6894, 'attacker_role': 'LWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Olivier Boscagli', 'market_value': 18500000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.6894}], 'xg_override': 0.0828}	aerialWon: 0.8464 -> 0.9 totalShots: 1.0709 -> 1.0612 shotOffTarget: 0.7345 -> 0.7278
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.0828 -> 0.0744 expected_assists: 0.0965 -> 0.0867
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 79.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0717` `goal_rate=0.0836` `scorer_weight=0.0748`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.083` `assist_rate=0.0768` `provider_weight=0.075`	-

Adam Wharton M 87 0 0.08 0.03 177% 1 0.7 7.1 57.6 29 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0125, quality 0.881)	`market_value_eur=79000000` `quality_factor=0.8811` `boost_factor=1.0125` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0096)	`blend_weight=0.0096` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=0.9654` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 2.6842 -> 2.663 totalShots: 1.095 -> 1.0845 interceptionWon: 0.9039 -> 0.9138 fouls: 0.7676 -> 0.7602 wasFouled: 1.4008 -> 1.4059
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yasin Ayari', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.522, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 1.1051, 'skill_edge': 0.0315, 'height_edge': 0.04, 'attacker_skill': 0.8877, 'defender_skill': 0.8177, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Yasin Ayari', 'market_value': 41000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8177}], 'xg_override': 0.0853}	aerialWon: 0.5624 -> 1.1 totalShots: 1.0913 -> 1.206 shotOffTarget: 0.7089 -> 0.7834 onTargetScoringAttempt: 0.3824 -> 0.4226 expectedGoals: 0.0772 -> 0.0853 goals: 0.0766 -> 0.0847
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.0853 -> 0.0766 expected_assists: 0.161 -> 0.1445
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 82.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0737` `goal_rate=0.0847` `scorer_weight=0.0763`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1345` `assist_rate=0.1252` `provider_weight=0.12`	-

Quinten Timber M 87 0 0.10 0.04 133% 1 0.7 6.9 31.1 9 XGBoost M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0273, quality 0.669)	`market_value_eur=24000000` `quality_factor=0.6685` `boost_factor=1.0273` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Maxim De Cuyper', 'matchup_zone': 'Right Wingback vs Left Back — Overlapping Run', 'duel_win_rate': 0.358, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v1', 'xg_multiplier': 0.9378, 'skill_edge': -0.0183, 'height_edge': -0.024, 'attacker_skill': 0.7474, 'defender_skill': 0.7881, 'attacker_role': 'RWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Maxim De Cuyper', 'market_value': 29000000, 'role': 'LB', 'price_index': 0.7, 'adjusted_quality': 0.7881}], 'xg_override': 0.1095}	aerialWon: 0.5965 -> 1.0 totalShots: 1.4565 -> 1.3659 shotOffTarget: 0.98 -> 0.919 onTargetScoringAttempt: 0.4764 -> 0.4468 goals: 0.1495 -> 0.1402 expectedGoals: 0.1168 -> 0.1095
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.1095 -> 0.0984 expected_assists: 0.0779 -> 0.07
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 79.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0937` `goal_rate=0.1402` `scorer_weight=0.1173`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0676` `assist_rate=0.083` `provider_weight=0.0728`	-

Anan Khalaili M 87 0 0.11 0.10 ok 1 0.8 6.9 41.8 12 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0223, quality 0.669)	`market_value_eur=24000000` `quality_factor=0.6685` `boost_factor=1.0223` `reference_quality=0.505` `history_matches=6` `history_shrink=0.4545`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0236)	`blend_weight=0.0236` `history_matches=6` `season_overlap_discount=1.0` `per90_to_match_factor=0.9706` `prior_reliability=0.08` `competition=Premier League`	wonTackle: 1.5802 -> 1.7149 totalCross: 2.853 -> 2.9194 interceptionWon: 1.1363 -> 1.1668 fouls: 0.853 -> 0.8711 wasFouled: 0.1439 -> 0.1596 totalShots: 1.3143 -> 1.302 onTargetScoringAttempt: 0.3629 -> 0.3543
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yasin Ayari', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.437, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9811, 'skill_edge': -0.0568, 'height_edge': 0.044, 'attacker_skill': 0.6914, 'defender_skill': 0.8177, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Yasin Ayari', 'market_value': 41000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8177}], 'xg_override': 0.122}	aerialWon: 1.4022 -> 1.1 totalShots: 1.3057 -> 1.281 shotOffTarget: 0.9514 -> 0.9334 bigChanceMissed: 0.3684 -> 0.3614 onTargetScoringAttempt: 0.3543 -> 0.3476
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.122 -> 0.1096 expected_assists: 0.1093 -> 0.0982
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 78.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1038` `goal_rate=0.1222` `scorer_weight=0.1097`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0935` `assist_rate=0.0931` `provider_weight=0.0882`	-

Jaydee Canvot D 90 0 0.04 0.05 ok 0 0.2 6.8 46.7 29 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0114, quality 0.732)	`market_value_eur=32000000` `quality_factor=0.732` `boost_factor=1.0114` `reference_quality=0.505` `history_matches=25` `history_shrink=0.1667`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0144)	`blend_weight=0.0144` `history_matches=25` `season_overlap_discount=1.0` `per90_to_match_factor=0.9948` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 1.2014 -> 1.2128 interceptionWon: 1.1661 -> 1.1564 fouls: 0.4866 -> 0.4796
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Pascal Groß', 'matchup_zone': 'LW vs CB'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 76.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0369` `goal_rate=0.0` `scorer_weight=0.0301`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0229` `assist_rate=0.0` `provider_weight=0.0186`	-

Chris Richards D 90 0 0.07 0.07 ok 1 0.5 6.8 53.6 28 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0059, quality 0.677)	`market_value_eur=25000000` `quality_factor=0.6773` `boost_factor=1.0059` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0098)	`blend_weight=0.0098` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9998` `prior_reliability=0.1333` `competition=Premier League`	interceptionWon: 1.1567 -> 1.165 wonTackle: 0.7031 -> 0.7109 fouls: 0.5663 -> 0.5607
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Charalampos Kostoulas', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.0811 -> 0.0729
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 80.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0703` `goal_rate=0.01` `scorer_weight=0.0583`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0206` `assist_rate=0.0` `provider_weight=0.0166`	-

Jefferson Lerma M 84 0 0.08 0.08 ok 1 0.6 6.8 46.5 26 FBref Prior M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9986, quality 0.468)	`market_value_eur=5500000` `quality_factor=0.4683` `boost_factor=0.9986` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Yasin Ayari', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.324, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.815, 'skill_edge': -0.1539, 'height_edge': 0.028, 'attacker_skill': 0.4758, 'defender_skill': 0.8177, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Yasin Ayari', 'market_value': 41000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8177}], 'xg_override': 0.0856}	aerialWon: 2.7156 -> 1.0 totalShots: 1.1082 -> 0.9032 shotOffTarget: 0.7749 -> 0.6315 onTargetScoringAttempt: 0.3333 -> 0.2716 bigChanceMissed: 0.2905 -> 0.2368 expectedGoalsOnTarget: 0.1638 -> 0.1335 expectedGoals: 0.105 -> 0.0856 goals: 0.0918 -> 0.0748
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	expected_goals: 0.0856 -> 0.0769 expected_assists: 0.0549 -> 0.0493
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 76.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.074` `goal_rate=0.0748` `scorer_weight=0.0708`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0481` `assist_rate=0.0549` `provider_weight=0.0493`	-

Walter Benítez G 90 0 0.00 - n/a 0.0 7.1 9.9 31 FBref Prior G_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9902, quality 0.446)	`market_value_eur=3700000` `quality_factor=0.4461` `boost_factor=0.9902` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Charalampos Kostoulas, Pascal Groß', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.4398}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.22 (squad xG sum 1.3534, xG scale 0.8984, xA scale 0.8984, team xA 0.81 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.216` `squad_xg_sum=1.3534` `squad_xa_sum=0.8965` `team_expected_assists=0.8055` `team_xa_to_xg_ratio=0.6624`	shots_faced: 13.7 -> 13.6 expected_saves: 3.36 -> 3.33 high_pressure_saves: 1.15 -> 1.14
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 78.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.2267`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.7186`	-

Team totals 1 1.22 12 6.9

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Dean Henderson	G	5.5m	15.4%	Market-value substitute prior
Honest Ahanor	D	14.9m	36.1%	Market-value substitute prior
Chadi Riad	D	14.3m	23.2%	Market-value substitute prior
Óscar Mingueza	M	19.0m	70.7%	Market-value substitute prior
Zavier Gozo	M	22.8m	70.1%	Market-value substitute prior
Ismaïla Sarr	F	26.7m	62.3%	Market-value substitute prior

Antagonist / tactical summary

Midfield is evenly contested — Yasin Ayari vs Adam Wharton will be the pivotal duel. Total antagonist xG — Home: 0.48, Away: 1.07. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Shadow Striker vs CDM + CB — Zone 14 Battle.

Midfield — possession 50.0% / 50.0%, dominance ratio 0.48, key duel: Yasin Ayari vs Adam Wharton

Pascal Groß xG 0.14Chema Andrés xG 0.09Charalampos Kostoulas xG 0.25Jørgen Strand Larsen xG 0.37Yéremy Pino xG 0.14Tyrick Mitchell xG 0.08

Summary

Antagonist matchup analysis for Brighton vs Crystal Palace. Most likely scoreline: 1-1. Lambdas derived from 3 home and 8 away individual player duels. Head-to-head: Brighton 2W-5D-3L vs Crystal Palace over 10 meetings.

Tactical parity between 4-2-3-1 and 3-4-2-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Yasin Ayari vs Adam Wharton will be the pivotal duel. Total antagonist xG — Home: 0.48, Away: 1.07. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Shadow Striker vs CDM + CB — Zone 14 Battle.

Leeds vs Man Utd

Premier League · 20261018 · 3-4-2-1 vs 4-2-3-1

Predicted 1-1Winner: DrawWinning H 1-0 (7.8%)Winning A 0-1 (9.8%)Confidence 0.72λ home/away 1.21 / 1.52Most-likely P 12.1%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Leeds vs Man Utd. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 3-4-2-1 vs 4-2-3-1 (duel lambdas: Home 1.281, Away 1.347). Final calibrated Poisson lambdas: Home 1.207, Away 1.524 (most likely score 1-1).

Home Starters 11/11 XGBoost (10 FBref) Away Starters 11/11 XGBoost (10 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.28 / 1.35 Final λ (H/A) 1.21 / 1.52

FBref Team Priors: Leeds United (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 1.724) vs Manchester United (Gls/90: 2.0, GA/90: 2.5 → E[λ]: 1.379) | Blend Weight: 0.192

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Leeds	Man Utd
League position	7.00	13.00
Points per game	1.42	1.17
Attack index	0.97	1.01
Defense (concede) index	1.01	0.85
Form multiplier	1.01	0.98
Standing λ multiplier	0.99	0.94

Lambda decomposition (how the goal rates were built)

Stage	Leeds λ	Man Utd λ
FBref Squad Prior E[λ] (weight 0.192)	1.724	1.379
Pre-competitive-floor	1.207	1.524
Pre-calibration (Micro-duels)	1.281	1.347
Final λ	1.207	1.524

Scoreline distribution & markets

Top 5 scorelines	P
1-1	12.1%
0-1	9.8%
1-2	9.1%
1-0	7.8%
0-2	7.6%

Home win	30.0%
Draw	25.5%
Away win	44.5%
Over 2.5	51.4%
BTTS yes	54.9%

Pregame fair line — favorite Man Utd (44.5%), fair total 2.73, fair spread -0.32

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Leeds — projected line vs season baseline (3-4-2-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Gabriel Gudmundsson M 84 0 0.04 0.01 298% 1 0.3 6.8 39.5 31 FBref Prior M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0066, quality 0.686)	`market_value_eur=26000000` `quality_factor=0.6858` `boost_factor=1.0066` `reference_quality=0.505` `history_matches=36` `history_shrink=0.122`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Lisandro Martínez', 'matchup_zone': 'Left Wingback vs Right Back — Overlapping Run', 'duel_win_rate': 0.377, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9667, 'skill_edge': -0.0466, 'height_edge': 0.024, 'attacker_skill': 0.765, 'defender_skill': 0.8686, 'attacker_role': 'LWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Lisandro Martínez', 'market_value': 48000000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.8686}], 'xg_override': 0.0405}	aerialWon: 0.7065 -> 1.1 totalShots: 0.691 -> 0.668 shotOffTarget: 0.5026 -> 0.4859 onTargetScoringAttempt: 0.1884 -> 0.1821
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 78.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0377` `goal_rate=0.0471` `scorer_weight=0.0412`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0873` `assist_rate=0.1009` `provider_weight=0.0972`	-

Ao Tanaka M 75 0 0.06 0.05 ok 1 0.4 6.8 35.3 31 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0034, quality 0.592)	`market_value_eur=16200000` `quality_factor=0.5921` `boost_factor=1.0034` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0028)	`blend_weight=0.0028` `history_matches=33` `season_overlap_discount=1.0` `per90_to_match_factor=0.8311` `prior_reliability=0.0333` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kobbie Mainoo', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.309, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.7923, 'skill_edge': -0.1253, 'height_edge': -0.016, 'attacker_skill': 0.6109, 'defender_skill': 0.8893, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Kobbie Mainoo', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}], 'xg_override': 0.0612}	aerialWon: 0.3241 -> 1.1 totalShots: 0.9809 -> 0.7772 shotOffTarget: 0.7065 -> 0.5598 onTargetScoringAttempt: 0.2744 -> 0.2174 bigChanceMissed: 0.1359 -> 0.1077 expectedGoals: 0.0772 -> 0.0612 goals: 0.0683 -> 0.0541 expectedGoalsOnTarget: 0.028 -> 0.0222
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 76.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0564` `goal_rate=0.0541` `scorer_weight=0.0543`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.063` `assist_rate=0.0638` `provider_weight=0.0642`	-

Ethan Ampadu M 90 0 0.05 0.05 ok 1 0.6 7.0 60.1 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0076, quality 0.739)	`market_value_eur=33000000` `quality_factor=0.7389` `boost_factor=1.0076` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0094)	`blend_weight=0.0094` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.9978` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 1.0257 -> 1.0207
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kobbie Mainoo', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.385, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9041, 'skill_edge': -0.0572, 'height_edge': -0.008, 'attacker_skill': 0.7622, 'defender_skill': 0.8893, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Kobbie Mainoo', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}], 'xg_override': 0.0484}	aerialWon: 1.3742 -> 1.1 totalShots: 0.7062 -> 0.6385 shotOffTarget: 0.5227 -> 0.4726 onTargetScoringAttempt: 0.1835 -> 0.1659 expectedGoalsOnTarget: 0.0992 -> 0.0897 goals: 0.0599 -> 0.0542 expectedGoals: 0.0535 -> 0.0484
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 80.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0449` `goal_rate=0.0542` `scorer_weight=0.0486`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0584` `assist_rate=0.0662` `provider_weight=0.0634`	-

Jayden Bogle M 79 0 0.13 0.08 52% 1 0.5 6.8 55.7 27 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0039, quality 0.623)	`market_value_eur=19200000` `quality_factor=0.6231` `boost_factor=1.0039` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0082)	`blend_weight=0.0082` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=0.873` `prior_reliability=0.1133` `competition=Premier League`	wasFouled: 1.1481 -> 1.1597 totalShots: 1.2309 -> 1.2208
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Luke Shaw', 'matchup_zone': 'Right Wingback vs Left Back — Overlapping Run', 'duel_win_rate': 0.475, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.1103, 'skill_edge': 0.0751, 'height_edge': 0.0, 'attacker_skill': 0.6976, 'defender_skill': 0.5308, 'attacker_role': 'RWB', 'attacker_mv_price_index': 0.7, 'defender_adjusted_parts': [{'player': 'Luke Shaw', 'market_value': 7500000, 'role': 'LB', 'price_index': 0.7, 'adjusted_quality': 0.5308}], 'xg_override': 0.1328}	totalShots: 1.2272 -> 1.3626 aerialWon: 0.8675 -> 1.0 shotOffTarget: 0.7782 -> 0.864 onTargetScoringAttempt: 0.449 -> 0.4985 expectedGoalsOnTarget: 0.2129 -> 0.2364 bigChanceMissed: 0.1896 -> 0.2105 goals: 0.1221 -> 0.1356 expectedGoals: 0.1196 -> 0.1328
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1328 -> 0.1258
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 76.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1182` `goal_rate=0.1356` `scorer_weight=0.1277`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0952` `assist_rate=0.0976` `provider_weight=0.0981`	-

Tarik Muharemović D 90 0 0.09 0.04 102% 1 0.5 7.0 38.7 11 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0396, quality 0.796)	`market_value_eur=43000000` `quality_factor=0.7957` `boost_factor=1.0396` `reference_quality=0.505` `history_matches=6` `history_shrink=0.4545`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0394)	`blend_weight=0.0394` `history_matches=6` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	interceptionWon: 1.2081 -> 1.2393 onTargetScoringAttempt: 0.1111 -> 0.1264 totalCross: 0.1307 -> 0.1452 wasFouled: 0.8 -> 0.7882 totalShots: 0.7193 -> 0.7107 wonTackle: 0.6868 -> 0.6794 goalAssist: 0.1809 -> 0.1738
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Bruno Fernandes', 'matchup_zone': 'LW vs CB'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0949 -> 0.0899
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 75.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.086` `goal_rate=0.0` `scorer_weight=0.0719`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0085` `assist_rate=0.1738` `provider_weight=0.0106`	-

Nico Elvedi D 89 0 0.03 0.01 161% 0 0.2 6.8 30.9 11 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0044, quality 0.534)	`market_value_eur=11000000` `quality_factor=0.5341` `boost_factor=1.0044` `reference_quality=0.505` `history_matches=5` `history_shrink=0.5`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0130)	`blend_weight=0.013` `history_matches=5` `season_overlap_discount=1.0` `per90_to_match_factor=0.9907` `prior_reliability=0.04` `competition=Premier League`	wasFouled: 0.0 -> 0.0429 wonTackle: 0.728 -> 0.7185 interceptionWon: 1.002 -> 1.0104 fouls: 0.5087 -> 0.5021 totalShots: 0.4926 -> 0.4862
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Matheus Cunha', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 78.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0275` `goal_rate=0.0` `scorer_weight=0.0223`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1082` `assist_rate=0.0` `provider_weight=0.0916`	-

James Justin D 88 0 0.08 0.06 ok 1 1.0 6.8 60.2 33 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0034, quality 0.595)	`market_value_eur=16500000` `quality_factor=0.5953` `boost_factor=1.0034` `reference_quality=0.505` `history_matches=35` `history_shrink=0.125`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0108)	`blend_weight=0.0108` `history_matches=35` `season_overlap_discount=1.0` `per90_to_match_factor=0.977` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 1.4355 -> 1.4464 totalShots: 1.0026 -> 0.997
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Amad Diallo', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 76.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0752` `goal_rate=0.2168` `scorer_weight=0.0978`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0634` `assist_rate=0.1158` `provider_weight=0.0819`	-

James Trafford G 90 0 0.00 - n/a 0.0 7.0 16.2 31 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0265, quality 0.770)	`market_value_eur=38000000` `quality_factor=0.7699` `boost_factor=1.0265` `reference_quality=0.505` `history_matches=10` `history_shrink=0.3333`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0289)	`blend_weight=0.0289` `history_matches=10` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Bryan Mbeumo, Bruno Fernandes', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.3467}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	shots_faced: 12.8 -> 14.5 expected_saves: 3.13 -> 3.55 high_pressure_saves: 1.08 -> 1.22
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 79.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.2764`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.7668`	-

Anton Stach M 87 0 0.14 0.10 36% 1 1.3 6.9 58.7 27 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0082, quality 0.725)	`market_value_eur=31000000` `quality_factor=0.7249` `boost_factor=1.0082` `reference_quality=0.505` `history_matches=35` `history_shrink=0.125`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0108)	`blend_weight=0.0108` `history_matches=35` `season_overlap_discount=1.0` `per90_to_match_factor=0.9703` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 6.7431 -> 6.7121 interceptionWon: 1.2786 -> 1.2963 wasFouled: 1.4004 -> 1.4168 fouls: 1.9183 -> 1.908 wonTackle: 1.5402 -> 1.534 totalShots: 1.4551 -> 1.4604
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kobbie Mainoo', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.419, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9538, 'skill_edge': -0.0634, 'height_edge': 0.032, 'attacker_skill': 0.7484, 'defender_skill': 0.8893, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Kobbie Mainoo', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}], 'xg_override': 0.1464}	aerialWon: 2.3909 -> 1.4 totalShots: 1.4604 -> 1.3929 shotOffTarget: 1.0103 -> 0.9636 onTargetScoringAttempt: 0.4451 -> 0.4245 bigChanceMissed: 0.2452 -> 0.2339 goals: 0.1632 -> 0.1557 expectedGoals: 0.1535 -> 0.1464
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1464 -> 0.1387
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 80.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1295` `goal_rate=0.1557` `scorer_weight=0.1453`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1037` `assist_rate=0.0942` `provider_weight=0.0987`	-

Noah Okafor F 72 0 0.15 0.12 ok 2 0.9 6.8 55.5 24 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0095, quality 0.752)	`market_value_eur=35000000` `quality_factor=0.752` `boost_factor=1.0095` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0056)	`blend_weight=0.0056` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.7951` `prior_reliability=0.0667` `competition=Premier League`	totalCross: 1.9046 -> 1.9161 wonTackle: 0.8229 -> 0.8316
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kobbie Mainoo, Harry Maguire', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle', 'duel_win_rate': 0.271, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 1.0013, 'skill_edge': 0.0169, 'height_edge': -0.016, 'attacker_skill': 0.752, 'defender_skill': 0.7144, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Kobbie Mainoo', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}, {'player': 'Harry Maguire', 'market_value': 8600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5395}], 'xg_override': 0.1544}	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1544 -> 0.1463
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 77.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1361` `goal_rate=0.1782` `scorer_weight=0.1633`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.063` `assist_rate=0.0493` `provider_weight=0.056`	-

Dominic Calvert-Lewin F 85 1 0.32 0.41 ok 2 1.7 6.9 55.8 30 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0062, quality 0.694)	`market_value_eur=27000000` `quality_factor=0.6941` `boost_factor=1.0062` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0094)	`blend_weight=0.0094` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.9494` `prior_reliability=0.1333` `competition=Premier League`	totalOffside: 0.6958 -> 0.7027 wasFouled: 1.0634 -> 1.0579
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kobbie Mainoo, Harry Maguire', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle', 'duel_win_rate': 0.253, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.9748, 'skill_edge': -0.0091, 'height_edge': -0.008, 'attacker_skill': 0.6941, 'defender_skill': 0.7144, 'attacker_role': 'SS', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Kobbie Mainoo', 'market_value': 77000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8893}, {'player': 'Harry Maguire', 'market_value': 8600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5395}], 'xg_override': 0.3385}	totalShots: 2.1068 -> 2.0537 shotOffTarget: 1.1976 -> 1.1674 onTargetScoringAttempt: 0.9071 -> 0.8842 bigChanceMissed: 0.7658 -> 0.7465 expectedGoalsOnTarget: 0.5302 -> 0.5168 goals: 0.3898 -> 0.38 expectedGoals: 0.3473 -> 0.3385
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.21 (squad xG sum 1.2741, xG scale 0.9474, xA scale 0.9960, team xA 0.78 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.207` `squad_xg_sum=1.2741` `squad_xa_sum=0.7877` `team_expected_assists=0.7846` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.3385 -> 0.3205
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 79.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.2742` `goal_rate=0.38` `scorer_weight=0.3547`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0285` `assist_rate=0.0042` `provider_weight=0.0231`	-

Team totals 1 1.21 13 6.9

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Lucas Perri	G	4.7m	12.2%	Market-value substitute prior
Joe Rodon	D	14.3m	23.2%	Market-value substitute prior
Jaka Bijol	D	14.3m	23.1%	Market-value substitute prior
Sean Longstaff	M	15.1m	70.3%	Market-value substitute prior
Ilia Gruev	M	22.4m	51.6%	Market-value substitute prior
Harry Wilson	F	36.6m	54.7%	Market-value substitute prior

Man Utd — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Senne Lammens G 90 0 0.00 - n/a 0.0 7.0 59.3 30 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0092, quality 0.770)	`market_value_eur=38000000` `quality_factor=0.7699` `boost_factor=1.0092` `reference_quality=0.505` `history_matches=38` `history_shrink=0.1163`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0101)	`blend_weight=0.0101` `history_matches=38` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Dominic Calvert-Lewin, Noah Okafor', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.2813}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	shots_faced: 12.2 -> 11.5 expected_saves: 2.99 -> 2.82 high_pressure_saves: 1.03 -> 0.97
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 82.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.6938`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=1.2581`	-

Bryan Mbeumo F 85 1 0.45 0.39 ok 2 2.2 7.1 57.5 33 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0126, quality 0.875)	`market_value_eur=73000000` `quality_factor=0.8747` `boost_factor=1.0126` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0098)	`blend_weight=0.0098` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9496` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 3.4363 -> 3.4726 totalShots: 2.3925 -> 2.411 interceptionWon: 0.6583 -> 0.6518
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Tarik Muharemović, Nico Elvedi', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.214, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 1.0202, 'skill_edge': 0.0737, 'height_edge': -0.06, 'attacker_skill': 0.8747, 'defender_skill': 0.7109, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Tarik Muharemović', 'market_value': 43000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8462}, {'player': 'Nico Elvedi', 'market_value': 11000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5755}], 'xg_override': 0.3853}	totalShots: 2.411 -> 2.4597 shotOffTarget: 1.3528 -> 1.3801 onTargetScoringAttempt: 1.0434 -> 1.0645 bigChanceMissed: 0.7171 -> 0.7316 expectedGoalsOnTarget: 0.4449 -> 0.4539 goals: 0.4185 -> 0.427 expectedGoals: 0.3777 -> 0.3853
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.3853 -> 0.4486 expected_assists: 0.1338 -> 0.1512
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 84.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.3615` `goal_rate=0.427` `scorer_weight=0.4916`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1403` `assist_rate=0.0254` `provider_weight=0.121`	-

Matheus Cunha F 84 0 0.21 0.11 80% 3 1.8 7.0 56.2 30 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0128, quality 0.881)	`market_value_eur=79000000` `quality_factor=0.8811` `boost_factor=1.0128` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0089)	`blend_weight=0.0089` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9371` `prior_reliability=0.12` `competition=Premier League`	wonTackle: 1.1126 -> 1.1027 fouls: 1.4114 -> 1.4035 interceptionWon: 0.5933 -> 0.588
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Tarik Muharemović, Nico Elvedi', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.251, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 1.0744, 'skill_edge': 0.0766, 'height_edge': -0.026, 'attacker_skill': 0.8811, 'defender_skill': 0.7109, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Tarik Muharemović', 'market_value': 43000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8462}, {'player': 'Nico Elvedi', 'market_value': 11000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5755}], 'xg_override': 0.1765}	totalShots: 2.1173 -> 2.2748 shotOffTarget: 1.4147 -> 1.52 onTargetScoringAttempt: 0.701 -> 0.7532 goals: 0.2198 -> 0.2362 expectedGoals: 0.1643 -> 0.1765 expectedGoalsOnTarget: 0.1623 -> 0.1744
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.1765 -> 0.2056 expected_assists: 0.1641 -> 0.1854
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 84.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1858` `goal_rate=0.2362` `scorer_weight=0.257`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1692` `assist_rate=0.2372` `provider_weight=0.2318`	-

Bruno Fernandes M 90 0 0.22 0.22 ok 2 2.0 7.1 63.0 29 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0074, quality 0.732)	`market_value_eur=32000000` `quality_factor=0.732` `boost_factor=1.0074` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0094)	`blend_weight=0.0094` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.9949` `prior_reliability=0.1333` `competition=Premier League`	totalShots: 2.3057 -> 2.3355 onTargetScoringAttempt: 0.7439 -> 0.7556 goals: 0.3442 -> 0.355 wasFouled: 0.7036 -> 0.697
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Tarik Muharemović', 'matchup_zone': 'LW vs CB', 'duel_win_rate': 0.313, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8422, 'skill_edge': -0.0514, 'height_edge': -0.056, 'attacker_skill': 0.732, 'defender_skill': 0.8462, 'attacker_role': 'LW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Tarik Muharemović', 'market_value': 43000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8462}], 'xg_override': 0.1866}	aerialWon: 0.2881 -> 1.0 totalShots: 2.3355 -> 1.967 shotOffTarget: 1.5618 -> 1.3153 onTargetScoringAttempt: 0.7556 -> 0.6364 goals: 0.355 -> 0.299 expectedGoalsOnTarget: 0.2576 -> 0.217 expectedGoals: 0.2216 -> 0.1866 bigChanceMissed: 0.1647 -> 0.1387
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.1866 -> 0.2174 expected_assists: 0.2272 -> 0.2567
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 89.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1954` `goal_rate=0.299` `scorer_weight=0.2717`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.2264` `assist_rate=0.3394` `provider_weight=0.3209`	-

Amad Diallo F 77 0 0.15 0.07 124% 2 1.0 6.8 43.9 29 XGBoost F_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0117, quality 0.800)	`market_value_eur=44000000` `quality_factor=0.8002` `boost_factor=1.0117` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Tarik Muharemović, Nico Elvedi', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.9709, 'skill_edge': 0.0402, 'height_edge': -0.06, 'attacker_skill': 0.8002, 'defender_skill': 0.7109, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Tarik Muharemović', 'market_value': 43000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8462}, {'player': 'Nico Elvedi', 'market_value': 11000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5755}], 'xg_override': 0.1319}	totalShots: 1.6802 -> 1.6313 shotOffTarget: 1.1023 -> 1.0702 onTargetScoringAttempt: 0.5779 -> 0.5611 bigChanceMissed: 0.1888 -> 0.1833 goals: 0.1872 -> 0.1818
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.1319 -> 0.1537 expected_assists: 0.0738 -> 0.0834
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 80.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1425` `goal_rate=0.1818` `scorer_weight=0.1921`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.08` `assist_rate=0.1416` `provider_weight=0.1043`	-

Kobbie Mainoo M 87 0 0.07 0.03 166% 1 0.6 7.3 58.7 28 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0144, quality 0.879)	`market_value_eur=77000000` `quality_factor=0.8792` `boost_factor=1.0144` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0061)	`blend_weight=0.0061` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.966` `prior_reliability=0.0733` `competition=Premier League`	totalShots: 0.8279 -> 0.8435 fouls: 0.928 -> 0.9223
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Noah Okafor, Dominic Calvert-Lewin, Anton Stach', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0579 -> 0.0675 expected_assists: 0.1004 -> 0.1135
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 81.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0653` `goal_rate=0.0581` `scorer_weight=0.0676`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1073` `assist_rate=0.0724` `provider_weight=0.0924`	-

Youri Tielemans M 80 0 0.08 0.05 58% 1 0.7 7.2 57.1 23 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0121, quality 0.796)	`market_value_eur=43000000` `quality_factor=0.7957` `boost_factor=1.0121` `reference_quality=0.505` `history_matches=31` `history_shrink=0.1389`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0102)	`blend_weight=0.0102` `history_matches=31` `season_overlap_discount=1.0` `per90_to_match_factor=0.8924` `prior_reliability=0.1133` `competition=Premier League`	wonTackle: 1.6271 -> 1.6158 totalShots: 0.9663 -> 0.9775 wasFouled: 2.0765 -> 2.0875 onTargetScoringAttempt: 0.2605 -> 0.2683 totalCross: 1.2703 -> 1.2627
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ethan Ampadu, Jayden Bogle', 'matchup_zone': 'Midfield lane'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0659 -> 0.0768 expected_assists: 0.1109 -> 0.1253
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 85.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0739` `goal_rate=0.0623` `scorer_weight=0.0744`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1178` `assist_rate=0.0751` `provider_weight=0.1002`	-

Luke Shaw D 85 0 0.02 0.02 ok 1 0.4 6.8 56.0 29 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9996, quality 0.493)	`market_value_eur=7500000` `quality_factor=0.4927` `boost_factor=0.9996` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0081)	`blend_weight=0.0081` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.9459` `prior_reliability=0.12` `competition=Premier League`	totalCross: 1.5215 -> 1.539 fouls: 0.9998 -> 0.9917
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jayden Bogle', 'matchup_zone': 'Right Wingback vs Left Back — Overlapping Run'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_assists: 0.0837 -> 0.0946
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 79.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0202` `goal_rate=0.059` `scorer_weight=0.0255`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0903` `assist_rate=0.1082` `provider_weight=0.1119`	-

Lisandro Martínez D 89 0 0.04 0.02 144% 1 0.5 7.1 51.9 22 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0167, quality 0.817)	`market_value_eur=48000000` `quality_factor=0.8168` `boost_factor=1.0167` `reference_quality=0.505` `history_matches=23` `history_shrink=0.1786`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0077)	`blend_weight=0.0077` `history_matches=23` `season_overlap_discount=1.0` `per90_to_match_factor=0.9851` `prior_reliability=0.0667` `competition=Premier League`	interceptionWon: 0.9356 -> 0.9284 wonTackle: 0.7999 -> 0.7937 fouls: 0.7732 -> 0.7672
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Gabriel Gudmundsson', 'matchup_zone': 'Left Wingback vs Right Back — Overlapping Run'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0322 -> 0.0375
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 82.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0368` `goal_rate=0.0` `scorer_weight=0.03`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.039` `assist_rate=0.0` `provider_weight=0.0318`	-

Harry Maguire D 89 0 0.06 0.04 54% 1 0.7 6.9 54.0 27 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0000, quality 0.506)	`market_value_eur=8600000` `quality_factor=0.5059` `boost_factor=1.0` `reference_quality=0.505` `history_matches=29` `history_shrink=0.1471`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0127)	`blend_weight=0.0127` `history_matches=29` `season_overlap_discount=1.0` `per90_to_match_factor=0.991` `prior_reliability=0.1333` `competition=Premier League`	wonTackle: 0.7684 -> 0.7586 totalCross: 0.506 -> 0.5122
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Noah Okafor', 'matchup_zone': 'Shadow Striker vs CDM + CB — Zone 14 Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0555 -> 0.0647 expected_assists: 0.05 -> 0.0565
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 82.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0627` `goal_rate=0.0044` `scorer_weight=0.0518`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0549` `assist_rate=0.0327` `provider_weight=0.0467`	-

Diogo Dalot D 88 0 0.08 0.04 128% 1 0.8 6.7 53.7 29 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0066, quality 0.702)	`market_value_eur=28000000` `quality_factor=0.7022` `boost_factor=1.0066` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0043)	`blend_weight=0.0043` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=0.9731` `prior_reliability=0.06` `competition=Premier League`	totalCross: 1.8427 -> 1.8581 totalShots: 0.7932 -> 0.8079
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Noah Okafor', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.52 (squad xG sum 1.3082, xG scale 1.1650, xA scale 1.1301, team xA 1.22 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.524` `squad_xg_sum=1.3082` `squad_xa_sum=1.0789` `team_expected_assists=1.2192` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0695 -> 0.081 expected_assists: 0.0658 -> 0.0744
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 16 by projected rating, fifa_overall 78.0)	`match_squad_rank=16` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0778` `goal_rate=0.003` `scorer_weight=0.0648`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0717` `assist_rate=0.0229` `provider_weight=0.0595`	-

Team totals 1 1.52 17 7.0

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Karl Darlow	G	4.4m	10.9%	Market-value substitute prior
Leny Yoro	D	16.1m	63.2%	Market-value substitute prior
Ayden Heaven	D	14.7m	24.1%	Market-value substitute prior
Carlos Baleba	M	23.6m	57.5%	Market-value substitute prior
Andrey Santos	M	16.6m	52.9%	Market-value substitute prior
Benjamin Šeško	F	22.1m	75.5%	Market-value substitute prior

Antagonist / tactical summary

Midfield is evenly contested — Ethan Ampadu vs Kobbie Mainoo will be the pivotal duel. Total antagonist xG — Home: 0.92, Away: 0.88. Home danger zone: Shadow Striker vs CDM + CB — Zone 14 Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Midfield — possession 49.1% / 50.9%, dominance ratio 0.47, key duel: Ethan Ampadu vs Kobbie Mainoo

Noah Okafor xG 0.15Dominic Calvert-Lewin xG 0.34Gabriel Gudmundsson xG 0.04Bruno Fernandes xG 0.19Matheus Cunha xG 0.18Bryan Mbeumo xG 0.39

Summary

Antagonist matchup analysis for Leeds vs Man Utd. Most likely scoreline: 1-1. Lambdas derived from 7 home and 4 away individual player duels. Head-to-head: Leeds 1W-4D-5L vs Man Utd over 10 meetings.

Tactical parity between 3-4-2-1 and 4-2-3-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Ethan Ampadu vs Kobbie Mainoo will be the pivotal duel. Total antagonist xG — Home: 0.92, Away: 0.88. Home danger zone: Shadow Striker vs CDM + CB — Zone 14 Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Málaga vs Real Sociedad

LaLiga · 20261018 · 4-1-4-1 vs 4-4-2

Predicted 0-1Winner: Real SociedadWinning H 1-0 (7.5%)Winning A 0-1 (17.4%)Confidence 0.76λ home/away 0.65 / 1.50Most-likely P 17.4%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Málaga vs Real Sociedad. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-1-4-1 vs 4-4-2 (duel lambdas: Home 0.696, Away 1.462). Final calibrated Poisson lambdas: Home 0.650, Away 1.505 (most likely score 0-1).

Home Starters 11/11 XGBoost (7 FBref) Away Starters 11/11 XGBoost (8 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 0.70 / 1.46 Final λ (H/A) 0.65 / 1.50

FBref Team Priors: Málaga CF (Gls/90: 0.3333333333333333, GA/90: 0.33 → E[λ]: 0.2) vs Real Sociedad (Gls/90: 0.75, GA/90: 0.75 → E[λ]: 0.2) | Blend Weight: 0.228

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Málaga	Real Sociedad
League position	20.00	6.00
Points per game	0.87	1.51
Attack index	0.73	0.99
Defense (concede) index	1.08	1.02
Form multiplier	0.97	1.03
Standing λ multiplier	0.91	1.01

Lambda decomposition (how the goal rates were built)

Stage	Málaga λ	Real Sociedad λ
FBref Squad Prior E[λ] (weight 0.228)	0.200	0.200
Pre-mismatch	0.701	1.452
Pre-competitive-floor	0.683	1.587
Pre-calibration (Micro-duels)	0.696	1.462
Final λ	0.650	1.505

competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Truecompetitive underdog floor applied Falsestage discount 1.0

Scoreline distribution & markets

Top 5 scorelines	P
0-1	17.4%
0-2	13.1%
0-0	11.7%
1-1	11.4%
1-2	8.5%

Home win	16.0%
Draw	26.2%
Away win	57.9%
Over 2.5	36.5%
BTTS yes	37.2%

Pregame fair line — favorite Real Sociedad (57.9%), fair total 2.15, fair spread -0.85

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Málaga — projected line vs season baseline (4-1-4-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Chupete F 88 0 0.20 0.31 34% 2 1.6 6.7 51.1 10 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0030, quality 0.531)	`market_value_eur=10700000` `quality_factor=0.5306` `boost_factor=1.003` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0500)	`blend_weight=0.05` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.9753` `prior_reliability=0.2` `competition=La Liga`	totalCross: 0.7862 -> 0.7469 totalShots: 2.0199 -> 2.0491 onTargetScoringAttempt: 0.8393 -> 0.8134 fouls: 1.2707 -> 1.2559 wonTackle: 0.37 -> 0.384 interceptionWon: 0.16 -> 0.152 wasFouled: 1.1096 -> 1.1029 totalOffside: 0.4557 -> 0.4492
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Mamadou Sarr, Héctor Fort', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.8755, 'skill_edge': -0.0667, 'height_edge': -0.018, 'attacker_skill': 0.5306, 'defender_skill': 0.6789, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Mamadou Sarr', 'market_value': 24000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.732}, {'player': 'Héctor Fort', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}], 'xg_override': 0.2644}	totalShots: 2.0491 -> 1.794 shotOffTarget: 1.1805 -> 1.0335 onTargetScoringAttempt: 0.8134 -> 0.7121 goals: 0.3418 -> 0.2992 expectedGoals: 0.302 -> 0.2644 bigChanceMissed: 0.2342 -> 0.205 expectedGoalsOnTarget: 0.233 -> 0.204
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.2644 -> 0.2019 expected_assists: 0.0343 -> 0.0262
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 80.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.1828` `goal_rate=0.2992` `scorer_weight=0.2243`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0259` `assist_rate=0.0` `provider_weight=0.021`	-

Adrián Niño F 80 0 0.14 0.10 39% 2 1.2 6.7 46.0 6 XGBoost F_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9887, quality 0.437)	`market_value_eur=1900000` `quality_factor=0.4374` `boost_factor=0.9887` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Mamadou Sarr, Héctor Fort', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.8197, 'skill_edge': -0.1087, 'height_edge': -0.014, 'attacker_skill': 0.4374, 'defender_skill': 0.6789, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Mamadou Sarr', 'market_value': 24000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.732}, {'player': 'Héctor Fort', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}], 'xg_override': 0.1864}	totalShots: 2.0559 -> 1.6852 shotOffTarget: 1.2583 -> 1.0314 onTargetScoringAttempt: 0.7976 -> 0.6538 goals: 0.2567 -> 0.2104 expectedGoals: 0.2274 -> 0.1864 expectedGoalsOnTarget: 0.2208 -> 0.181
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.1864 -> 0.1424 expected_assists: 0.034 -> 0.026
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 14 by projected rating, fifa_overall 72.0)	`match_squad_rank=14` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.1327` `goal_rate=0.2104` `scorer_weight=0.1569`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0257` `assist_rate=0.0` `provider_weight=0.0208`	-

Dani Lorenzo M 79 0 0.03 0.03 ok 0 0.2 6.9 53.7 9 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9916, quality 0.437)	`market_value_eur=2300000` `quality_factor=0.4374` `boost_factor=0.9916` `reference_quality=0.505` `history_matches=7` `history_shrink=0.4167`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0325)	`blend_weight=0.0325` `history_matches=7` `season_overlap_discount=1.0` `per90_to_match_factor=0.8738` `prior_reliability=0.12` `competition=La Liga`	fouls: 0.9023 -> 0.873 totalShots: 0.6115 -> 0.5916 wonTackle: 1.0937 -> 1.0739 interceptionWon: 0.6232 -> 0.6345 totalCross: 0.203 -> 0.1964 onTargetScoringAttempt: 0.1997 -> 0.1932
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Jon Gorrotxategi', 'matchup_zone': 'Central Mid vs Central Mid — Box-to-Box Midfield Contest', 'duel_win_rate': 0.29, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v1', 'xg_multiplier': 0.8088, 'skill_edge': -0.1261, 'height_edge': -0.004, 'attacker_skill': 0.4374, 'defender_skill': 0.7176, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Jon Gorrotxategi', 'market_value': 27000000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.7176}], 'xg_override': 0.0362}	aerialWon: 0.2388 -> 0.9 totalShots: 0.6049 -> 0.4893 shotOffTarget: 0.4117 -> 0.333 onTargetScoringAttempt: 0.1932 -> 0.1563 bigChanceMissed: 0.1844 -> 0.1491 goals: 0.0592 -> 0.0479 expectedGoals: 0.0447 -> 0.0362
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.0362 -> 0.0277 expected_assists: 0.0638 -> 0.0487
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 72.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0273` `goal_rate=0.0479` `scorer_weight=0.0315`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0475` `assist_rate=0.0597` `provider_weight=0.0469`	-

David Larrubia F 88 0 0.06 0.07 ok 1 1.4 7.1 52.2 18 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9969, quality 0.478)	`market_value_eur=6300000` `quality_factor=0.4781` `boost_factor=0.9969` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0400)	`blend_weight=0.04` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.9753` `prior_reliability=0.16` `competition=La Liga`	wonTackle: 1.2478 -> 1.3117 totalShots: 1.7307 -> 1.7758 totalCross: 1.0694 -> 1.0266 fouls: 1.0994 -> 1.0717 wasFouled: 1.3211 -> 1.3008 totalOffside: 0.0977 -> 0.11 onTargetScoringAttempt: 0.5412 -> 0.5523 goals: 0.1418 -> 0.1361
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Mamadou Sarr, Héctor Fort', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.779, 'skill_edge': -0.0904, 'height_edge': -0.06, 'attacker_skill': 0.4781, 'defender_skill': 0.6789, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Mamadou Sarr', 'market_value': 24000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.732}, {'player': 'Héctor Fort', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}], 'xg_override': 0.0848}	totalShots: 1.7758 -> 1.3833 shotOffTarget: 1.1895 -> 0.9266 onTargetScoringAttempt: 0.5523 -> 0.4302 goals: 0.1361 -> 0.106 expectedGoals: 0.1089 -> 0.0848 expectedGoalsOnTarget: 0.0858 -> 0.0668
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.0848 -> 0.0648 expected_assists: 0.1595 -> 0.1219
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 81.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0627` `goal_rate=0.106` `scorer_weight=0.075`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.1148` `assist_rate=0.0` `provider_weight=0.0975`	-

Carlos Dotor M 81 0 0.05 0.03 49% 1 0.4 6.9 55.0 10 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9916, quality 0.437)	`market_value_eur=1200000` `quality_factor=0.4374` `boost_factor=0.9916` `reference_quality=0.505` `history_matches=7` `history_shrink=0.4167`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0451)	`blend_weight=0.0451` `history_matches=7` `season_overlap_discount=1.0` `per90_to_match_factor=0.8953` `prior_reliability=0.1667` `competition=La Liga`	wonTackle: 1.0809 -> 1.1291 fouls: 1.2684 -> 1.2273 wasFouled: 0.6806 -> 0.6499 totalCross: 0.1246 -> 0.1513 totalShots: 0.7418 -> 0.7241
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Jon Gorrotxategi', 'matchup_zone': 'Holding Mid vs Central Mid — Midfield Anchor Duel', 'duel_win_rate': 0.286, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v1', 'xg_multiplier': 0.8323, 'skill_edge': -0.1261, 'height_edge': 0.012, 'attacker_skill': 0.4374, 'defender_skill': 0.7176, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Jon Gorrotxategi', 'market_value': 27000000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.7176}], 'xg_override': 0.0653}	aerialWon: 0.3451 -> 1.1 totalShots: 0.7466 -> 0.6214 shotOffTarget: 0.5 -> 0.4162 onTargetScoringAttempt: 0.2466 -> 0.2052 expectedGoalsOnTarget: 0.2403 -> 0.2 goals: 0.0811 -> 0.0675 expectedGoals: 0.0785 -> 0.0653
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.0653 -> 0.0499 expected_assists: 0.0707 -> 0.054
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 85.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0487` `goal_rate=0.0675` `scorer_weight=0.0509`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0526` `assist_rate=0.0697` `provider_weight=0.0536`	-

Jose Salinas D 83 0 0.00 0.01 51% 0 0.1 6.6 18.5 30 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9954, quality 0.437)	`market_value_eur=1800000` `quality_factor=0.4374` `boost_factor=0.9954` `reference_quality=0.505` `history_matches=17` `history_shrink=0.2273`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0059)	`blend_weight=0.0059` `history_matches=17` `season_overlap_discount=1.0` `per90_to_match_factor=0.9189` `prior_reliability=0.04` `competition=La Liga`	fouls: 1.5985 -> 1.5891 totalCross: 1.1364 -> 1.1297 wasFouled: 0.9784 -> 0.9726 wonTackle: 0.9564 -> 0.9507
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Takefusa Kubo', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_assists: 0.0388 -> 0.0296
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 18 by projected rating, fifa_overall 77.0)	`match_squad_rank=18` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0028` `goal_rate=0.0` `scorer_weight=0.0026`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0292` `assist_rate=0.0` `provider_weight=0.0237`	-

Ángel Recio D 90 0 0.02 0.01 ok 1 0.6 6.9 54.4 10 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9922, quality 0.437)	`market_value_eur=1000000` `quality_factor=0.4374` `boost_factor=0.9922` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Mikel Oyarzabal', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 65.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0159` `goal_rate=0.0` `scorer_weight=0.0128`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0064` `assist_rate=0.0` `provider_weight=0.0055`	-

Fernando Calero D 85 0 0.03 0.03 ok 0 0.3 6.8 53.7 20 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9970, quality 0.437)	`market_value_eur=2100000` `quality_factor=0.4374` `boost_factor=0.997` `reference_quality=0.505` `history_matches=29` `history_shrink=0.1471`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Mikel Oyarzabal', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.0388 -> 0.0296
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 81.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0292` `goal_rate=0.0` `scorer_weight=0.0237`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0099` `assist_rate=0.0` `provider_weight=0.0079`	-

Rafa Garrido D 88 0 0.00 0.00 37% 0 0.1 6.7 53.8 10 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9922, quality 0.437)	`market_value_eur=1400000` `quality_factor=0.4374` `boost_factor=0.9922` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Carlos Soler', 'matchup_zone': 'Left Mid vs Right Back — Wide Channel'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_assists: 0.0262 -> 0.02
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 79.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0019` `goal_rate=0.0` `scorer_weight=0.0018`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0198` `assist_rate=0.2191` `provider_weight=0.025`	-

Alfonso Herrero G 90 0 0.00 - n/a 0.0 7.1 58.7 10 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9922, quality 0.437)	`market_value_eur=420000` `quality_factor=0.4374` `boost_factor=0.9922` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0500)	`blend_weight=0.05` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2` `competition=La Liga`	saves: 2.9467 -> 2.9994 wasFouled: 0.2995 -> 0.2845 fouls: 0.1478 -> 0.1404
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Orri Óskarsson, Mikel Oyarzabal', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.4617}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	shots_faced: 13.9 -> 14.3 expected_saves: 3.4 -> 3.5 high_pressure_saves: 1.17 -> 1.2
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 75.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=0.6936`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.4548`	-

Pablo Martínez M 81 0 0.05 0.05 ok 1 0.8 6.9 60.5 26 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9974, quality 0.439)	`market_value_eur=3100000` `quality_factor=0.4387` `boost_factor=0.9974` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0028)	`blend_weight=0.0028` `history_matches=33` `season_overlap_discount=1.0` `per90_to_match_factor=0.8977` `prior_reliability=0.0333` `competition=La Liga`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Jon Martin', 'matchup_zone': 'Left Mid vs Right Back — Wide Channel', 'duel_win_rate': 0.311, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v1', 'xg_multiplier': 0.781, 'skill_edge': -0.137, 'height_edge': -0.012, 'attacker_skill': 0.4429, 'defender_skill': 0.7474, 'attacker_role': 'LM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Jon Martin', 'market_value': 24000000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.7474}], 'xg_override': 0.0656}	aerialWon: 0.4074 -> 1.1 totalShots: 1.0863 -> 0.8484 shotOffTarget: 0.7262 -> 0.5672 onTargetScoringAttempt: 0.3601 -> 0.2812 expectedGoals: 0.084 -> 0.0656 goals: 0.0767 -> 0.0599
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 0 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=0`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.65 (squad xG sum 0.8509, xG scale 0.7639, xA scale 0.7639, team xA 0.50 = 0.77 x lambda)	`scoreline_reset=true` `team_lambda=0.65` `squad_xg_sum=0.8509` `squad_xa_sum=0.6556` `team_expected_assists=0.5008` `team_xa_to_xg_ratio=0.7705`	expected_goals: 0.0656 -> 0.0501 expected_assists: 0.1447 -> 0.1105
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 82.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation skipped: 0 projected goals	`team_goals=0` `allocation_applied=false` `goals_allocated=0` `goal_probability=0.0489` `goal_rate=0.0599` `scorer_weight=0.0476`	-
`integer_assist_projection`	Team scoreline allocation skipped: 0 projected assists	`allocation_applied=false` `team_assists=0` `assists_allocated=0` `assist_probability=0.1046` `assist_rate=0.1472` `provider_weight=0.1119`	-

Team totals 0 0.65 10 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Carlos Lopez	G	4.4m	10.9%	Market-value substitute prior
Adam Aznou	D	13.7m	30.6%	Market-value substitute prior
Diego Murillo	D	13.6m	29.9%	Market-value substitute prior
Jens-Lys Cajuste	M	22.1m	50.7%	Market-value substitute prior
Izan Merino	M	22.1m	50.3%	Market-value substitute prior
Joaquín Muñoz	F	25.3m	55.9%	Market-value substitute prior

Real Sociedad — projected line vs season baseline (4-4-2)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Mikel Oyarzabal F 79 0 0.29 0.34 ok 2 1.6 7.0 63.5 32 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0060, quality 0.694)	`market_value_eur=27000000` `quality_factor=0.6941` `boost_factor=1.006` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0138)	`blend_weight=0.0138` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.8763` `prior_reliability=0.2` `competition=La Liga`	wasFouled: 1.1752 -> 1.1872 wonTackle: 0.8094 -> 0.7982 onTargetScoringAttempt: 0.7327 -> 0.7388 goalAssist: 0.1648 -> 0.1706
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ángel Recio, Fernando Calero', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.304, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.1522, 'skill_edge': 0.1155, 'height_edge': -0.012, 'attacker_skill': 0.6941, 'defender_skill': 0.4374, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ángel Recio', 'market_value': 1000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}, {'player': 'Fernando Calero', 'market_value': 2100000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.2999}	totalShots: 2.0356 -> 2.3454 shotOffTarget: 1.2968 -> 1.4942 onTargetScoringAttempt: 0.7388 -> 0.8512 bigChanceMissed: 0.6712 -> 0.7734 expectedGoals: 0.2603 -> 0.2999 goals: 0.2566 -> 0.2957 expectedGoalsOnTarget: 0.2492 -> 0.2871
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.2999 -> 0.2939 expected_assists: 0.0988 -> 0.1352
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 90.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.2546` `goal_rate=0.2957` `scorer_weight=0.2904`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1265` `assist_rate=0.1706` `provider_weight=0.169`	-

Ander Barrenetxea F 70 0 0.14 0.06 137% 2 0.9 6.8 55.0 32 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0035, quality 0.603)	`market_value_eur=17200000` `quality_factor=0.6027` `boost_factor=1.0035` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0124)	`blend_weight=0.0124` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.7772` `prior_reliability=0.16` `competition=La Liga`	totalCross: 2.543 -> 2.5356 interceptionWon: 0.5372 -> 0.5305
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ángel Recio, Fernando Calero', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.234, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.0505, 'skill_edge': 0.0744, 'height_edge': -0.04, 'attacker_skill': 0.6027, 'defender_skill': 0.4374, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ángel Recio', 'market_value': 1000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}, {'player': 'Fernando Calero', 'market_value': 2100000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.1468}	totalShots: 1.5038 -> 1.5798 shotOffTarget: 1.0189 -> 1.0704 onTargetScoringAttempt: 0.4849 -> 0.5094 goals: 0.2127 -> 0.2234 expectedGoals: 0.1397 -> 0.1468 expectedGoalsOnTarget: 0.1282 -> 0.1347
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_assists: 0.1045 -> 0.143
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 81.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1339` `goal_rate=0.2234` `scorer_weight=0.1798`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1332` `assist_rate=0.2415` `provider_weight=0.1787`	-

Orri Óskarsson F 78 1 0.40 0.29 40% 3 1.2 6.8 55.6 27 XGBoost F_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0066, quality 0.641)	`market_value_eur=21000000` `quality_factor=0.6408` `boost_factor=1.0066` `reference_quality=0.505` `history_matches=26` `history_shrink=0.1613`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ángel Recio, Fernando Calero', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.3, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.1463, 'skill_edge': 0.0915, 'height_edge': 0.008, 'attacker_skill': 0.6408, 'defender_skill': 0.4374, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ángel Recio', 'market_value': 1000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}, {'player': 'Fernando Calero', 'market_value': 2100000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.412}	totalShots: 2.2235 -> 2.5488 shotOffTarget: 1.2435 -> 1.4254 onTargetScoringAttempt: 0.98 -> 1.1234 bigChanceMissed: 0.9177 -> 1.052 expectedGoalsOnTarget: 0.7178 -> 0.8228 goals: 0.4466 -> 0.5119 expectedGoals: 0.3594 -> 0.412
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	goals: 1.0 -> 0
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.412 -> 0.4036 expected_assists: 0.0304 -> 0.0416
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 74.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.3321` `goal_rate=0.5119` `scorer_weight=0.4909`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0407` `assist_rate=0.0` `provider_weight=0.0333`	-

Carlos Soler M 83 0 0.12 0.09 35% 1 0.9 6.9 63.2 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9992, quality 0.482)	`market_value_eur=6600000` `quality_factor=0.4818` `boost_factor=0.9992` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0108)	`blend_weight=0.0108` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.9261` `prior_reliability=0.1533` `competition=La Liga`	totalShots: 1.3162 -> 1.3063 wonTackle: 0.3337 -> 0.3432 fouls: 0.5346 -> 0.5288
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Rafa Garrido', 'matchup_zone': 'Left Mid vs Right Back — Wide Channel', 'duel_win_rate': 0.524, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.094, 'skill_edge': 0.024, 'height_edge': 0.04, 'attacker_skill': 0.4907, 'defender_skill': 0.4374, 'attacker_role': 'LM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Rafa Garrido', 'market_value': 1400000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.4374}], 'xg_override': 0.1248}	aerialWon: 0.1438 -> 1.1 totalShots: 1.3118 -> 1.4352 shotOffTarget: 0.9144 -> 1.0004 onTargetScoringAttempt: 0.3974 -> 0.4348 expectedGoalsOnTarget: 0.1518 -> 0.1661 goals: 0.1184 -> 0.1295 expectedGoals: 0.1141 -> 0.1248 bigChanceMissed: 0.086 -> 0.0941
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_assists: 0.1379 -> 0.1887
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 79.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1151` `goal_rate=0.1295` `scorer_weight=0.1256`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.172` `assist_rate=0.1242` `provider_weight=0.1749`	-

Jon Gorrotxategi M 74 0 0.04 0.01 435% 1 0.2 6.9 50.5 24 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0071, quality 0.694)	`market_value_eur=27000000` `quality_factor=0.6941` `boost_factor=1.0071` `reference_quality=0.505` `history_matches=35` `history_shrink=0.125`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0016)	`blend_weight=0.0016` `history_matches=35` `season_overlap_discount=1.0` `per90_to_match_factor=0.8214` `prior_reliability=0.02` `competition=La Liga`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Carlos Dotor', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.564, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.1677, 'skill_edge': 0.1261, 'height_edge': -0.012, 'attacker_skill': 0.7176, 'defender_skill': 0.4374, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Carlos Dotor', 'market_value': 1200000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.4374}], 'xg_override': 0.0404}	aerialWon: 1.1942 -> 1.0 totalShots: 0.5056 -> 0.5904 shotOffTarget: 0.3832 -> 0.4475 onTargetScoringAttempt: 0.1224 -> 0.1429 goals: 0.0463 -> 0.0541 expectedGoals: 0.0346 -> 0.0404
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_assists: 0.038 -> 0.052
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 79.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0388` `goal_rate=0.0541` `scorer_weight=0.0456`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0507` `assist_rate=0.0473` `provider_weight=0.0575`	-

Takefusa Kubo F 70 0 0.16 0.05 201% 2 0.4 6.8 43.0 27 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0060, quality 0.641)	`market_value_eur=21000000` `quality_factor=0.6408` `boost_factor=1.006` `reference_quality=0.505` `history_matches=29` `history_shrink=0.1471`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0147)	`blend_weight=0.0147` `history_matches=29` `season_overlap_discount=1.0` `per90_to_match_factor=0.7751` `prior_reliability=0.1533` `competition=La Liga`	wasFouled: 2.2211 -> 2.2132 onTargetScoringAttempt: 0.4805 -> 0.4735 totalShots: 1.0987 -> 1.0927
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ángel Recio, Fernando Calero', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.252, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.0757, 'skill_edge': 0.0915, 'height_edge': -0.04, 'attacker_skill': 0.6408, 'defender_skill': 0.4374, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Ángel Recio', 'market_value': 1000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}, {'player': 'Fernando Calero', 'market_value': 2100000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.1615}	totalShots: 1.0927 -> 1.1754 shotOffTarget: 0.6182 -> 0.665 onTargetScoringAttempt: 0.4735 -> 0.5093 bigChanceMissed: 0.1989 -> 0.214 goals: 0.169 -> 0.1818 expectedGoals: 0.1501 -> 0.1615
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_assists: 0.0887 -> 0.1214
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 80.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1463` `goal_rate=0.1818` `scorer_weight=0.1734`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1143` `assist_rate=0.0967` `provider_weight=0.1284`	-

Sergio Gómez D 86 0 0.02 0.02 ok 1 0.4 6.8 60.8 29 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0022, quality 0.575)	`market_value_eur=14600000` `quality_factor=0.5748` `boost_factor=1.0022` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0172)	`blend_weight=0.0172` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.9511` `prior_reliability=0.2533` `competition=La Liga`	totalCross: 4.4367 -> 4.4593
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Adrián Niño', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_assists: 0.1155 -> 0.158
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 78.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0154` `goal_rate=0.0` `scorer_weight=0.0124`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1462` `assist_rate=0.1365` `provider_weight=0.1781`	-

Jon Martin D 90 0 0.09 0.06 38% 1 0.4 6.8 53.8 29 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0066, quality 0.669)	`market_value_eur=24000000` `quality_factor=0.6685` `boost_factor=1.0066` `reference_quality=0.505` `history_matches=32` `history_shrink=0.1351`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0234)	`blend_weight=0.0234` `history_matches=32` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2667` `competition=La Liga`	interceptionWon: 1.1107 -> 1.0905 wonTackle: 0.792 -> 0.8086 wasFouled: 0.4635 -> 0.4702 goals: 0.2568 -> 0.2508
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Pablo Martínez', 'matchup_zone': 'Left Mid vs Right Back — Wide Channel'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 78.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0844` `goal_rate=0.2508` `scorer_weight=0.1103`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0122` `assist_rate=0.0` `provider_weight=0.0104`	-

Mamadou Sarr D 85 0 0.08 0.02 248% 0 0.1 6.7 22.0 18 FBref Prior D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0223, quality 0.669)	`market_value_eur=24000000` `quality_factor=0.6685` `boost_factor=1.0223` `reference_quality=0.505` `history_matches=6` `history_shrink=0.4545`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Chupete', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 16 by projected rating, fifa_overall 77.0)	`match_squad_rank=16` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0808` `goal_rate=0.0` `scorer_weight=0.0674`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0107` `assist_rate=0.0` `provider_weight=0.0093`	-

Héctor Fort D 63 0 0.06 0.00 1211% 1 0.1 6.6 33.2 14 FBref Prior D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0050, quality 0.575)	`market_value_eur=14600000` `quality_factor=0.5748` `boost_factor=1.005` `reference_quality=0.505` `history_matches=16` `history_shrink=0.2381`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Chupete', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	expected_assists: 0.0172 -> 0.0235
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 18 by projected rating, fifa_overall 72.0)	`match_squad_rank=18` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0627` `goal_rate=0.3822` `scorer_weight=0.0809`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0232` `assist_rate=0.2726` `provider_weight=0.0294`	-

Álex Remiro G 90 0 0.00 - n/a 0.0 6.9 63.7 30 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0004, quality 0.518)	`market_value_eur=9600000` `quality_factor=0.5177` `boost_factor=1.0004` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0177)	`blend_weight=0.0177` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2667` `competition=La Liga`	saves: 2.9674 -> 2.9901
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Chupete, Adrián Niño', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 0.6959}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.50 (squad xG sum 1.5360, xG scale 0.9798, xA scale 1.3683, team xA 0.98 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.505` `squad_xg_sum=1.536` `squad_xa_sum=0.7149` `team_expected_assists=0.9782` `team_xa_to_xg_ratio=0.65`	shots_faced: 6.6 -> 6.2 expected_saves: 1.61 -> 1.52 high_pressure_saves: 0.56 -> 0.52
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 81.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.678`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=1.0475`	-

Team totals 1 1.50 16 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Unai Marrero	G	4.4m	10.9%	Market-value substitute prior
Jon Aramburu	D	14.1m	22.7%	Market-value substitute prior
Igor Zubeldia	D	24.9m	31.6%	Market-value substitute prior
Luka Sučić	M	26.7m	41.6%	Market-value substitute prior
Beñat Turrientes	M	20.5m	41.1%	Market-value substitute prior
Gonçalo Guedes	F	20.8m	72.2%	Market-value substitute prior

Antagonist / tactical summary

Away midfield dictates play with ~52% possession expected (Pablo Martínez vs Jon Gorrotxategi). Total antagonist xG — Home: 0.70, Away: 1.19. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Midfield — possession 47.5% / 52.5%, dominance ratio 0.44, key duel: Pablo Martínez vs Jon Gorrotxategi

Chupete xG 0.26Adrián Niño xG 0.19David Larrubia xG 0.08Carlos Soler xG 0.12Jon Gorrotxategi xG 0.04Mikel Oyarzabal xG 0.30

Summary

Antagonist matchup analysis for Málaga vs Real Sociedad. Most likely scoreline: 0-1. Lambdas derived from 6 home and 6 away individual player duels. Head-to-head: Málaga 4W-3D-3L vs Real Sociedad over 10 meetings.

Tactical parity between 4-1-4-1 and 4-4-2. Match outcome driven primarily by individual player quality. Away midfield dictates play with ~52% possession expected (Pablo Martínez vs Jon Gorrotxategi). Total antagonist xG — Home: 0.70, Away: 1.19. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Osasuna vs Racing Club

LaLiga · 20261018 · 4-3-3 vs 4-2-3-1

Predicted 1-1Winner: DrawWinning H 1-0 (8.7%)Winning A 0-1 (9.3%)Confidence 0.7λ home/away 1.30 / 1.39Most-likely P 12.4%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Osasuna vs Racing Club. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-3-3 vs 4-2-3-1 (duel lambdas: Home 1.343, Away 1.351). Final calibrated Poisson lambdas: Home 1.305, Away 1.390 (most likely score 1-1).

Home Starters 11/11 XGBoost (9 FBref) Away Starters 11/11 XGBoost (8 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.34 / 1.35 Final λ (H/A) 1.30 / 1.39

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Osasuna	Racing Club
League position	16.00	15.00
Points per game	1.16	1.16
Attack index	0.86	1.02
Defense (concede) index	1.34	1.18
Form multiplier	1.00	1.00
Standing λ multiplier	1.06	1.08

Lambda decomposition (how the goal rates were built)

Stage	Osasuna λ	Racing Club λ
Pre-competitive-floor	1.305	1.390
Pre-calibration (Micro-duels)	1.343	1.351
Final λ	1.305	1.390

Scoreline distribution & markets

Top 5 scorelines	P
1-1	12.4%
0-1	9.3%
1-0	8.7%
1-2	8.5%
2-1	8.0%

Home win	35.0%
Draw	26.1%
Away win	38.9%
Over 2.5	50.5%
BTTS yes	54.9%

Pregame fair line — favorite Racing Club (38.9%), fair total 2.69, fair spread -0.09

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Osasuna — projected line vs season baseline (4-3-3)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Aitor Fernández G 90 0 0.00 - n/a 0.0 6.9 11.5 31 FBref Prior G_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9899, quality 0.437)	`market_value_eur=375000` `quality_factor=0.4374` `boost_factor=0.9899` `reference_quality=0.505` `history_matches=5` `history_shrink=0.5`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Yassir Zabiri, Pablo García', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.3505}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	shots_faced: 12.8 -> 13.2 expected_saves: 3.13 -> 3.23 high_pressure_saves: 1.08 -> 1.11
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 77.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.3652`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.9792`	-

Moi Gómez M 78 0 0.05 0.01 850% 1 0.1 7.1 45.4 29 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9972, quality 0.437)	`market_value_eur=1600000` `quality_factor=0.4374` `boost_factor=0.9972` `reference_quality=0.505` `history_matches=31` `history_shrink=0.1389`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0120)	`blend_weight=0.012` `history_matches=31` `season_overlap_discount=1.0` `per90_to_match_factor=0.8721` `prior_reliability=0.1333` `competition=La Liga`	totalCross: 1.6902 -> 1.6751 wonTackle: 0.6211 -> 0.6136 totalShots: 0.5816 -> 0.5746
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Íñigo Sáinz-Maza', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.41, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9412, 'skill_edge': 0.0, 'height_edge': -0.04, 'attacker_skill': 0.4374, 'defender_skill': 0.4374, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Íñigo Sáinz-Maza', 'market_value': 1000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.4374}], 'xg_override': 0.0349}	aerialWon: 0.5907 -> 0.9 totalShots: 0.5799 -> 0.5458 shotOffTarget: 0.4371 -> 0.4114 onTargetScoringAttempt: 0.1428 -> 0.1344
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0349 -> 0.0511 expected_assists: 0.1097 -> 0.1606
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 76.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0498` `goal_rate=0.0456` `scorer_weight=0.0575`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1484` `assist_rate=0.0906` `provider_weight=0.1414`	-

Ante Budimir F 87 1 0.60 0.49 ok 3 2.7 6.8 65.5 31 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9980, quality 0.437)	`market_value_eur=2600000` `quality_factor=0.4374` `boost_factor=0.998` `reference_quality=0.505` `history_matches=45` `history_shrink=0.1`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0095)	`blend_weight=0.0095` `history_matches=45` `season_overlap_discount=1.0` `per90_to_match_factor=0.9619` `prior_reliability=0.1467` `competition=La Liga`	totalShots: 2.7238 -> 2.7346 wasFouled: 1.5669 -> 1.577 onTargetScoringAttempt: 1.1767 -> 1.1859 wonTackle: 0.4576 -> 0.4657
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Jeanuël Belocian, Alvaro Mantilla', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.187, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9806, 'skill_edge': -0.0372, 'height_edge': 0.024, 'attacker_skill': 0.4374, 'defender_skill': 0.52, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Jeanuël Belocian', 'market_value': 12900000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6027}, {'player': 'Alvaro Mantilla', 'market_value': 1200000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.4115}	totalShots: 2.7346 -> 2.6815 shotOffTarget: 1.5471 -> 1.5171 onTargetScoringAttempt: 1.1859 -> 1.1629 bigChanceMissed: 0.8206 -> 0.8047 expectedGoalsOnTarget: 0.659 -> 0.6462 goals: 0.4576 -> 0.4487 expectedGoals: 0.4196 -> 0.4115
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.4115 -> 0.6025 expected_assists: 0.0331 -> 0.0485
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 82.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.4526` `goal_rate=0.4487` `scorer_weight=0.6511`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0473` `assist_rate=0.0` `provider_weight=0.0388`	-

Rubén García F 70 0 0.15 0.07 125% 1 0.8 6.8 57.4 40 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9979, quality 0.437)	`market_value_eur=1700000` `quality_factor=0.4374` `boost_factor=0.9979` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0077)	`blend_weight=0.0077` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.7731` `prior_reliability=0.1133` `competition=La Liga`	totalCross: 4.1566 -> 4.1666 totalShots: 1.2032 -> 1.2083
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Pablo Ramón', 'matchup_zone': 'Left Wing vs Right Back — Flank Battle', 'duel_win_rate': 0.42, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9118, 'skill_edge': 0.0, 'height_edge': -0.06, 'attacker_skill': 0.4374, 'defender_skill': 0.4374, 'attacker_role': 'LW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Pablo Ramón', 'market_value': 1500000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.4374}], 'xg_override': 0.1007}	totalShots: 1.2083 -> 1.1017 shotOffTarget: 0.8102 -> 0.7387 onTargetScoringAttempt: 0.3972 -> 0.3622 bigChanceMissed: 0.1599 -> 0.1458 goals: 0.1507 -> 0.1374 expectedGoals: 0.1104 -> 0.1007 expectedGoalsOnTarget: 0.0598 -> 0.0545
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.1007 -> 0.1474 expected_assists: 0.0775 -> 0.1135
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 77.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1371` `goal_rate=0.1374` `scorer_weight=0.1833`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1073` `assist_rate=0.1602` `provider_weight=0.1419`	-

Iker Muñoz M 83 0 0.05 0.01 284% 1 0.3 6.9 44.9 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9967, quality 0.437)	`market_value_eur=1800000` `quality_factor=0.4374` `boost_factor=0.9967` `reference_quality=0.505` `history_matches=26` `history_shrink=0.1613`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0161)	`blend_weight=0.0161` `history_matches=26` `season_overlap_discount=1.0` `per90_to_match_factor=0.9236` `prior_reliability=0.1533` `competition=La Liga`	totalCross: 0.4221 -> 0.4347 wasFouled: 1.4883 -> 1.4773 fouls: 1.4334 -> 1.4426 interceptionWon: 1.2159 -> 1.2093
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Íñigo Sáinz-Maza', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.4, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 0.9706, 'skill_edge': 0.0, 'height_edge': -0.02, 'attacker_skill': 0.4374, 'defender_skill': 0.4374, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Íñigo Sáinz-Maza', 'market_value': 1000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.4374}], 'xg_override': 0.0347}	aerialWon: 1.7725 -> 1.0 totalShots: 0.6132 -> 0.5952 shotOffTarget: 0.4576 -> 0.4441
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0347 -> 0.0508 expected_assists: 0.0413 -> 0.0605
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 73.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0495` `goal_rate=0.0399` `scorer_weight=0.0539`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0587` `assist_rate=0.058` `provider_weight=0.0716`	-

Lucas Torró M 78 0 0.13 0.03 277% 1 0.7 6.8 48.4 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9977, quality 0.437)	`market_value_eur=2700000` `quality_factor=0.4374` `boost_factor=0.9977` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0063)	`blend_weight=0.0063` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=0.8718` `prior_reliability=0.0867` `competition=La Liga`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Íñigo Sáinz-Maza', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.474, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.0353, 'skill_edge': 0.0, 'height_edge': 0.024, 'attacker_skill': 0.4374, 'defender_skill': 0.4374, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Íñigo Sáinz-Maza', 'market_value': 1000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.4374}], 'xg_override': 0.0879}	aerialWon: 3.2092 -> 1.3 totalShots: 0.9178 -> 0.9502 shotOffTarget: 0.5771 -> 0.5975 onTargetScoringAttempt: 0.3403 -> 0.3523
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0879 -> 0.1287 expected_assists: 0.0379 -> 0.0555
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 76.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1208` `goal_rate=0.0741` `scorer_weight=0.1158`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.054` `assist_rate=0.0486` `provider_weight=0.0623`	-

Jon Moncayola M 78 0 0.06 0.05 ok 1 0.6 6.8 62.4 29 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9995, quality 0.490)	`market_value_eur=7300000` `quality_factor=0.4903` `boost_factor=0.9995` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0111)	`blend_weight=0.0111` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=0.8702` `prior_reliability=0.1667` `competition=La Liga`	totalCross: 2.6467 -> 2.629 interceptionWon: 0.786 -> 0.7966 wonTackle: 1.2202 -> 1.226
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Íñigo Sáinz-Maza', 'matchup_zone': 'Central Mid vs Defensive Mid — Pressing/Creation Duel', 'duel_win_rate': 0.47, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.0296, 'skill_edge': 0.0282, 'height_edge': -0.008, 'attacker_skill': 0.5, 'defender_skill': 0.4374, 'attacker_role': 'CM', 'attacker_mv_price_index': 0.9, 'defender_adjusted_parts': [{'player': 'Íñigo Sáinz-Maza', 'market_value': 1000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.4374}], 'xg_override': 0.0437}	aerialWon: 0.5966 -> 1.1 totalShots: 0.626 -> 0.6445 shotOffTarget: 0.4464 -> 0.4596 onTargetScoringAttempt: 0.1796 -> 0.1849
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0437 -> 0.064 expected_assists: 0.0798 -> 0.1168
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 81.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.062` `goal_rate=0.0494` `scorer_weight=0.0679`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1102` `assist_rate=0.0775` `provider_weight=0.1147`	-

Abel Bretones D 88 0 0.02 0.00 1989% 0 0.1 6.6 45.2 30 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9977, quality 0.442)	`market_value_eur=3400000` `quality_factor=0.4424` `boost_factor=0.9977` `reference_quality=0.505` `history_matches=36` `history_shrink=0.122`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0159)	`blend_weight=0.0159` `history_matches=36` `season_overlap_discount=1.0` `per90_to_match_factor=0.9827` `prior_reliability=0.2` `competition=La Liga`	wonTackle: 1.1419 -> 1.155 fouls: 1.1349 -> 1.1429 wasFouled: 0.8333 -> 0.8253
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Iñigo Vicente', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.016 -> 0.0234 expected_assists: 0.0773 -> 0.1132
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 16 by projected rating, fifa_overall 76.0)	`match_squad_rank=16` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0231` `goal_rate=0.0104` `scorer_weight=0.0214`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.107` `assist_rate=0.0484` `provider_weight=0.0906`	-

Enzo Boyomo D 85 0 0.05 0.03 100% 0 0.3 6.8 43.3 26 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0028, quality 0.575)	`market_value_eur=14600000` `quality_factor=0.5748` `boost_factor=1.0028` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0034)	`blend_weight=0.0034` `history_matches=33` `season_overlap_discount=1.0` `per90_to_match_factor=0.9449` `prior_reliability=0.04` `competition=La Liga`	wonTackle: 0.8913 -> 0.899 fouls: 1.1814 -> 1.1881
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Yassir Zabiri', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0366 -> 0.0536 expected_assists: 0.0116 -> 0.017
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 79.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0522` `goal_rate=0.0096` `scorer_weight=0.0429`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0169` `assist_rate=0.0927` `provider_weight=0.0213`	-

Alejandro Catena D 89 0 0.02 0.03 ok 0 0.3 6.9 61.9 29 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9978, quality 0.437)	`market_value_eur=2700000` `quality_factor=0.4374` `boost_factor=0.9978` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Yassir Zabiri', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0166 -> 0.0243 expected_assists: 0.1309 -> 0.1917
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 85.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.024` `goal_rate=0.0248` `scorer_weight=0.0273`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1744` `assist_rate=0.1552` `provider_weight=0.2174`	-

Iñigo Arguibide D 78 0 0.02 0.00 571% 0 0.1 6.6 20.8 31 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9947, quality 0.437)	`market_value_eur=2700000` `quality_factor=0.4374` `boost_factor=0.9947` `reference_quality=0.505` `history_matches=14` `history_shrink=0.2632`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0331)	`blend_weight=0.0331` `history_matches=14` `season_overlap_discount=1.0` `per90_to_match_factor=0.8689` `prior_reliability=0.1933` `competition=La Liga`	totalCross: 2.2688 -> 2.3028 wonTackle: 0.9588 -> 0.9865 wasFouled: 1.0779 -> 1.0522 fouls: 1.2714 -> 1.2492 totalShots: 0.4585 -> 0.4534
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Sergio Canales', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.8913, xG scale 1.4642, xA scale 1.4642, team xA 0.98 = 0.75 x lambda)	`scoreline_reset=true` `team_lambda=1.305` `squad_xg_sum=0.8913` `squad_xa_sum=0.6669` `team_expected_assists=0.9765` `team_xa_to_xg_ratio=0.7482`	expected_goals: 0.0122 -> 0.0179 expected_assists: 0.0242 -> 0.0354
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 18 by projected rating, fifa_overall 81.0)	`match_squad_rank=18` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0177` `goal_rate=0.0` `scorer_weight=0.0144`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0348` `assist_rate=0.0` `provider_weight=0.0283`	-

Team totals 1 1.30 10 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Sergio Herrera	G	4.4m	13.1%	Market-value substitute prior
Jorge Herrando	D	13.6m	21.7%	Market-value substitute prior
Valentin Rosier	D	13.6m	29.9%	Market-value substitute prior
Aimar Oroz	M	14.9m	43.9%	Market-value substitute prior
Mauro Echegoyen	M	22.0m	49.9%	Market-value substitute prior
Romain Del Castillo	F	25.5m	56.6%	Market-value substitute prior

Racing Club — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Yassir Zabiri F 80 1 0.56 0.36 56% 3 1.5 6.7 52.5 11 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0168, quality 0.650)	`market_value_eur=22000000` `quality_factor=0.6503` `boost_factor=1.0168` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0217)	`blend_weight=0.0217` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.8863` `prior_reliability=0.0867` `competition=La Liga`	goals: 0.5215 -> 0.5545 totalOffside: 0.3412 -> 0.3634 totalShots: 2.4976 -> 2.5174 onTargetScoringAttempt: 1.2298 -> 1.2475 wasFouled: 0.5995 -> 0.5865 wonTackle: 0.5348 -> 0.5232 interceptionWon: 0.2331 -> 0.228
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Enzo Boyomo, Alejandro Catena', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.23, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.0437, 'skill_edge': 0.0517, 'height_edge': -0.022, 'attacker_skill': 0.6503, 'defender_skill': 0.5353, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Enzo Boyomo', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}, {'player': 'Alejandro Catena', 'market_value': 2700000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4449}], 'xg_override': 0.4541}	totalShots: 2.5174 -> 2.6274 shotOffTarget: 1.2677 -> 1.3231 onTargetScoringAttempt: 1.2475 -> 1.302 expectedGoalsOnTarget: 0.7812 -> 0.8153 goals: 0.5545 -> 0.5787 expectedGoals: 0.4351 -> 0.4541
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	goals: 1.0 -> 0
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.4541 -> 0.5614 expected_assists: 0.0369 -> 0.0456
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 66.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.4296` `goal_rate=0.5787` `scorer_weight=0.7002`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0446` `assist_rate=0.0` `provider_weight=0.0365`	-

Iñigo Vicente F 85 0 0.09 0.03 201% 1 0.4 7.0 48.5 21 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9973, quality 0.482)	`market_value_eur=6600000` `quality_factor=0.4818` `boost_factor=0.9973` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0467)	`blend_weight=0.0467` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.9461` `prior_reliability=0.1867` `competition=La Liga`	totalCross: 4.945 -> 4.8404 wonTackle: 1.4141 -> 1.3796 totalShots: 1.0342 -> 1.0018 wasFouled: 2.6454 -> 2.6166 onTargetScoringAttempt: 0.3862 -> 0.3682 fouls: 1.2263 -> 1.2164 interceptionWon: 0.5206 -> 0.5121 goalAssist: 0.4942 -> 0.4869
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Enzo Boyomo, Alejandro Catena', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9087, 'skill_edge': -0.0241, 'height_edge': -0.038, 'attacker_skill': 0.4818, 'defender_skill': 0.5353, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Enzo Boyomo', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}, {'player': 'Alejandro Catena', 'market_value': 2700000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4449}], 'xg_override': 0.0731}	totalShots: 1.0162 -> 0.9234 shotOffTarget: 0.648 -> 0.5888 onTargetScoringAttempt: 0.3682 -> 0.3346 goals: 0.1155 -> 0.105 expectedGoals: 0.0804 -> 0.0731
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.0731 -> 0.0904 expected_assists: 0.184 -> 0.2273
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 77.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0864` `goal_rate=0.105` `scorer_weight=0.113`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.2033` `assist_rate=0.4869` `provider_weight=0.2841`	-

Pablo García F 74 0 0.18 0.06 177% 2 0.8 6.8 33.0 28 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9998, quality 0.501)	`market_value_eur=8199999` `quality_factor=0.5011` `boost_factor=0.9998` `reference_quality=0.505` `history_matches=22` `history_shrink=0.1852`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0048)	`blend_weight=0.0048` `history_matches=22` `season_overlap_discount=1.0` `per90_to_match_factor=0.8197` `prior_reliability=0.04` `competition=La Liga`	wasFouled: 1.0313 -> 1.0526 fouls: 0.4718 -> 0.4893 totalShots: 1.909 -> 1.9195 goals: 0.2033 -> 0.2089 totalCross: 1.1399 -> 1.1344
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Enzo Boyomo, Alejandro Catena', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.9038, 'skill_edge': -0.0154, 'height_edge': -0.05, 'attacker_skill': 0.5011, 'defender_skill': 0.5353, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Enzo Boyomo', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}, {'player': 'Alejandro Catena', 'market_value': 2700000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4449}], 'xg_override': 0.1443}	totalShots: 1.9195 -> 1.7348 shotOffTarget: 1.2487 -> 1.1286 onTargetScoringAttempt: 0.6637 -> 0.5999 goals: 0.2089 -> 0.1888 bigChanceMissed: 0.1948 -> 0.1761 expectedGoalsOnTarget: 0.1786 -> 0.1614 expectedGoals: 0.1597 -> 0.1443
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.1443 -> 0.1784 expected_assists: 0.0611 -> 0.0755
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 70.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1634` `goal_rate=0.1888` `scorer_weight=0.2193`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0727` `assist_rate=0.0` `provider_weight=0.0604`	-

Sergio Canales F 80 0 0.13 0.10 37% 1 1.6 6.9 51.9 19 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9925, quality 0.440)	`market_value_eur=3200000` `quality_factor=0.4399` `boost_factor=0.9925` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0417)	`blend_weight=0.0417` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.8883` `prior_reliability=0.1667` `competition=La Liga`	totalCross: 2.5581 -> 2.4663 totalShots: 1.4989 -> 1.5708 wonTackle: 0.7958 -> 0.7626 onTargetScoringAttempt: 0.5016 -> 0.5255 interceptionWon: 0.512 -> 0.4907 wasFouled: 1.4717 -> 1.4548 fouls: 0.5855 -> 0.5759 goals: 0.1244 -> 0.1192
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Enzo Boyomo, Alejandro Catena', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.8693, 'skill_edge': -0.0429, 'height_edge': -0.046, 'attacker_skill': 0.4399, 'defender_skill': 0.5353, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Enzo Boyomo', 'market_value': 14600000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6258}, {'player': 'Alejandro Catena', 'market_value': 2700000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4449}], 'xg_override': 0.1077}	totalShots: 1.5708 -> 1.3655 shotOffTarget: 0.9973 -> 0.867 onTargetScoringAttempt: 0.5255 -> 0.4568 expectedGoalsOnTarget: 0.2014 -> 0.1751 expectedGoals: 0.1239 -> 0.1077 goals: 0.1192 -> 0.1036
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.1077 -> 0.1331 expected_assists: 0.0887 -> 0.1097
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 78.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1246` `goal_rate=0.1036` `scorer_weight=0.1296`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1039` `assist_rate=0.0` `provider_weight=0.0878`	-

Íñigo Sáinz-Maza M 83 0 0.05 - n/a 1 0.0 6.7 11.7 11 FBref Prior M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9855, quality 0.437)	`market_value_eur=1000000` `quality_factor=0.4374` `boost_factor=0.9855` `reference_quality=0.505` `history_matches=2` `history_shrink=0.7143`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Moi Gómez', 'matchup_zone': 'Holding Mid vs Central Mid — Midfield Anchor Duel', 'duel_win_rate': 0.44, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.0588, 'skill_edge': 0.0, 'height_edge': 0.04, 'attacker_skill': 0.4374, 'defender_skill': 0.4374, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Moi Gómez', 'market_value': 1600000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.4374}], 'xg_override': 0.0366}	aerialWon: 0.0 -> 1.1 totalShots: 0.5565 -> 0.5892 shotOffTarget: 0.4174 -> 0.4419 onTargetScoringAttempt: 0.1391 -> 0.1473
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.0366 -> 0.0452 expected_assists: 0.0418 -> 0.0517
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 14 by projected rating, fifa_overall 67.0)	`match_squad_rank=14` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0442` `goal_rate=0.052` `scorer_weight=0.0532`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0504` `assist_rate=0.0488` `provider_weight=0.0556`	-

Iván Martín M 86 0 0.06 0.03 113% 1 0.4 7.0 57.8 31 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9978, quality 0.437)	`market_value_eur=2700000` `quality_factor=0.4374` `boost_factor=0.9978` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0071)	`blend_weight=0.0071` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.9503` `prior_reliability=0.1` `competition=La Liga`	wonTackle: 0.9642 -> 0.9753 totalCross: 0.3919 -> 0.3981
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Moi Gómez', 'matchup_zone': 'Holding Mid vs Central Mid — Midfield Anchor Duel', 'duel_win_rate': 0.417, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.025, 'skill_edge': 0.001, 'height_edge': 0.016, 'attacker_skill': 0.4396, 'defender_skill': 0.4374, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Moi Gómez', 'market_value': 1600000, 'role': 'CM', 'price_index': 0.9, 'adjusted_quality': 0.4374}], 'xg_override': 0.0487}	aerialWon: 1.0742 -> 1.0 totalShots: 0.6846 -> 0.7017 shotOffTarget: 0.4466 -> 0.4578 onTargetScoringAttempt: 0.2366 -> 0.2425
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.0487 -> 0.0602 expected_assists: 0.0824 -> 0.1019
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 77.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0584` `goal_rate=0.082` `scorer_weight=0.0752`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0969` `assist_rate=0.067` `provider_weight=0.09`	-

Jorge Salinas D 90 0 0.07 0.03 114% 1 0.4 6.7 47.5 11 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0000, quality 0.505)	`market_value_eur=8500000` `quality_factor=0.5047` `boost_factor=1.0` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0500)	`blend_weight=0.05` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2` `competition=La Liga`	wasFouled: 1.4307 -> 1.4092 totalCross: 1.3315 -> 1.3483 fouls: 0.9738 -> 0.9584 interceptionWon: 1.2391 -> 1.2271 totalShots: 0.5563 -> 0.545 goalAssist: 0.1595 -> 0.1515
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=true`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Rubén García', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.0602 -> 0.0744 expected_assists: 0.1429 -> 0.1767
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 12 by projected rating, fifa_overall 72.0)	`match_squad_rank=12` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0717` `goal_rate=0.0` `scorer_weight=0.0595`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.162` `assist_rate=0.1515` `provider_weight=0.1846`	-

Jeanuël Belocian D 90 0 0.05 0.02 97% 1 0.4 6.7 32.9 8 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0085, quality 0.556)	`market_value_eur=12900000` `quality_factor=0.5559` `boost_factor=1.0085` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ante Budimir', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_goals: 0.0382 -> 0.0472
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 71.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0461` `goal_rate=0.0` `scorer_weight=0.0378`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0146` `assist_rate=0.0` `provider_weight=0.0119`	-

Pablo Ramón D 83 0 0.01 0.00 207% 1 0.2 6.8 36.8 11 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9908, quality 0.437)	`market_value_eur=1500000` `quality_factor=0.4374` `boost_factor=0.9908` `reference_quality=0.505` `history_matches=6` `history_shrink=0.4545`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0355)	`blend_weight=0.0355` `history_matches=6` `season_overlap_discount=1.0` `per90_to_match_factor=0.9234` `prior_reliability=0.12` `competition=La Liga`	wonTackle: 1.0723 -> 1.107 interceptionWon: 1.1453 -> 1.1775 totalShots: 0.538 -> 0.5189 onTargetScoringAttempt: 0.2271 -> 0.219
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Rubén García', 'matchup_zone': 'Left Wing vs Right Back — Flank Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 69.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0082` `goal_rate=0.0` `scorer_weight=0.0072`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0105` `assist_rate=0.0` `provider_weight=0.009`	-

Alvaro Mantilla D 86 0 0.02 0.02 ok 0 0.2 6.7 40.2 11 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9916, quality 0.437)	`market_value_eur=1200000` `quality_factor=0.4374` `boost_factor=0.9916` `reference_quality=0.505` `history_matches=7` `history_shrink=0.4167`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ante Budimir', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	expected_assists: 0.0231 -> 0.0286
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 15 by projected rating, fifa_overall 77.0)	`match_squad_rank=15` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0176` `goal_rate=0.0` `scorer_weight=0.0142`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0282` `assist_rate=0.0` `provider_weight=0.0229`	-

Julen Agirrezabala G 90 0 0.00 - n/a 0.0 6.9 37.3 16 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9988, quality 0.481)	`market_value_eur=6500000` `quality_factor=0.4805` `boost_factor=0.9988` `reference_quality=0.505` `history_matches=26` `history_shrink=0.1613`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0210)	`blend_weight=0.021` `history_matches=26` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2` `competition=La Liga`	saves: 2.6138 -> 2.6289 wasFouled: 0.1259 -> 0.1372
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ante Budimir, Rubén García', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.3425}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.39 (squad xG sum 1.1244, xG scale 1.2362, xA scale 1.2362, team xA 0.92 = 0.66 x lambda)	`scoreline_reset=true` `team_lambda=1.39` `squad_xg_sum=1.1244` `squad_xa_sum=0.7429` `team_expected_assists=0.9183` `team_xa_to_xg_ratio=0.6607`	shots_faced: 12.8 -> 12.4 expected_saves: 3.14 -> 3.04 high_pressure_saves: 1.07 -> 1.04
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 82.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.5727`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.9036`	-

Team totals 1 1.39 15 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Simon Eriksson	G	4.4m	10.9%	Market-value substitute prior
Facundo González	D	13.6m	29.9%	Market-value substitute prior
Pedro Felipe	D	13.6m	29.9%	Market-value substitute prior
Gustavo Puerta	M	22.7m	32.2%	Market-value substitute prior
André Almeida	M	22.1m	50.6%	Market-value substitute prior
Andrés Martín	F	25.5m	57.0%	Market-value substitute prior

Antagonist / tactical summary

Midfield is evenly contested — Jon Moncayola vs Iván Martín will be the pivotal duel. Total antagonist xG — Home: 0.71, Away: 0.86. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Midfield — possession 50.7% / 49.3%, dominance ratio 0.54, key duel: Jon Moncayola vs Iván Martín

Rubén García xG 0.10Ante Budimir xG 0.41Iker Muñoz xG 0.03Yassir Zabiri xG 0.45Iñigo Vicente xG 0.07Pablo García xG 0.14

Summary

Antagonist matchup analysis for Osasuna vs Racing Club. Most likely scoreline: 1-1. Lambdas derived from 6 home and 6 away individual player duels. Head-to-head: Osasuna 3W-0D-7L vs Racing Club over 10 meetings.

Tactical parity between 4-3-3 and 4-2-3-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Jon Moncayola vs Iván Martín will be the pivotal duel. Total antagonist xG — Home: 0.71, Away: 0.86. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Real Madrid vs Sevilla

LaLiga · 20261018 · 4-2-3-1 vs 4-2-3-1

Predicted 3-1Winner: Real MadridWinning H 3-1 (7.6%)Winning A 1-2 (2.3%)Confidence 0.82λ home/away 3.57 / 1.30Most-likely P 7.6%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Real Madrid vs Sevilla. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-2-3-1 vs 4-2-3-1 (duel lambdas: Home 3.272, Away 1.470). Final calibrated Poisson lambdas: Home 3.569, Away 1.300 (most likely score 3-1).

Home Starters 11/11 XGBoost (10 FBref) Away Starters 11/11 XGBoost (7 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 3.27 / 1.47 Final λ (H/A) 3.57 / 1.30

FBref Team Priors: Real Madrid (Gls/90: 3.0, GA/90: 3.33 → E[λ]: 3.5) vs Sevilla (Gls/90: 2.0, GA/90: 2.0 → E[λ]: 3.5) | Blend Weight: 0.228

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Real Madrid	Sevilla
League position	4.00	5.00
Points per game	1.87	1.59
Attack index	1.22	1.02
Defense (concede) index	0.99	0.86
Form multiplier	1.02	0.98
Standing λ multiplier	1.08	0.95

Lambda decomposition (how the goal rates were built)

Stage	Real Madrid λ	Sevilla λ
FBref Squad Prior E[λ] (weight 0.228)	3.500	3.500
Pre-mismatch	3.333	1.536
Pre-calibration (Micro-duels)	3.272	1.470
Final λ	3.569	1.300

Scoreline distribution & markets

Top 5 scorelines	P
3-1	7.6%
4-1	6.8%
2-1	6.4%
3-0	5.8%
4-0	5.2%

Home win	79.0%
Draw	11.4%
Away win	9.6%
Over 2.5	86.3%
BTTS yes	70.7%

Pregame fair line — favorite Real Madrid (79.0%), fair total 4.87, fair spread 2.27

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Real Madrid — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Kylian Mbappé F 89 2 1.47 0.71 107% 5 4.2 7.4 64.5 24 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0135, quality 0.900)	`market_value_eur=212000000` `quality_factor=0.9` `boost_factor=1.0135` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0148)	`blend_weight=0.0148` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9904` `prior_reliability=0.2` `competition=La Liga`	totalShots: 4.028 -> 4.0855 onTargetScoringAttempt: 1.9465 -> 1.9861 totalOffside: 1.6717 -> 1.6519 goals: 0.6007 -> 0.6113 totalCross: 1.3619 -> 1.3515 wasFouled: 1.3699 -> 1.3643
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Andrés Castrin, Kike Salas', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.312, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.1648, 'skill_edge': 0.1401, 'height_edge': -0.028, 'attacker_skill': 0.9, 'defender_skill': 0.5887, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Andrés Castrin', 'market_value': 8700000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5411}, {'player': 'Kike Salas', 'market_value': 15400000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6363}], 'xg_override': 0.7852}	totalShots: 4.0855 -> 4.7588 shotOffTarget: 2.0814 -> 2.4244 onTargetScoringAttempt: 1.9861 -> 2.3134 bigChanceMissed: 1.3341 -> 1.554 expectedGoalsOnTarget: 0.8274 -> 0.9638 expectedGoals: 0.6741 -> 0.7852 goals: 0.6113 -> 0.712
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	goals: 1.0 -> 0
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.7852 -> 1.4662 expected_assists: 0.1626 -> 0.3918
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 91.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=2` `goal_probability=0.7692` `goal_rate=0.712` `scorer_weight=1.3378`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=1` `assist_probability=0.3242` `assist_rate=0.21` `provider_weight=0.4792`	-

Yan Diomande F 82 0 0.32 0.06 402% 2 0.7 6.9 60.5 13 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0454, quality 0.899)	`market_value_eur=132000000` `quality_factor=0.8986` `boost_factor=1.0454` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0117)	`blend_weight=0.0117` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.9093` `prior_reliability=0.0467` `competition=La Liga`	wasFouled: 2.3234 -> 2.2963 totalCross: 3.5425 -> 3.5618 wonTackle: 1.5717 -> 1.5534 totalShots: 1.4226 -> 1.406 fouls: 0.9845 -> 0.973 interceptionWon: 0.9326 -> 0.9217 onTargetScoringAttempt: 0.5418 -> 0.5355
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Andrés Castrin, Kike Salas', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.311, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.1639, 'skill_edge': 0.1395, 'height_edge': -0.028, 'attacker_skill': 0.8986, 'defender_skill': 0.5887, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Andrés Castrin', 'market_value': 8700000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.5411}, {'player': 'Kike Salas', 'market_value': 15400000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6363}], 'xg_override': 0.1726}	totalShots: 1.4162 -> 1.6483 shotOffTarget: 0.8807 -> 1.025 onTargetScoringAttempt: 0.5355 -> 0.6233 expectedGoalsOnTarget: 0.3342 -> 0.389 goals: 0.1728 -> 0.2011 expectedGoals: 0.1483 -> 0.1726
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1726 -> 0.3223 expected_assists: 0.1311 -> 0.3159
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 84.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.2755` `goal_rate=0.2011` `scorer_weight=0.3636`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.2709` `assist_rate=0.3414` `provider_weight=0.3949`	-

Jude Bellingham M 84 1 0.44 0.17 159% 3 1.6 7.1 66.8 24 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0144, quality 0.899)	`market_value_eur=151000000` `quality_factor=0.8995` `boost_factor=1.0144` `reference_quality=0.505` `history_matches=36` `history_shrink=0.122`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0143)	`blend_weight=0.0143` `history_matches=36` `season_overlap_discount=1.0` `per90_to_match_factor=0.9348` `prior_reliability=0.18` `competition=La Liga`	totalShots: 1.9174 -> 1.9298 onTargetScoringAttempt: 0.7271 -> 0.7366 goalAssist: 0.1183 -> 0.1265 wasFouled: 2.9921 -> 2.984 goals: 0.1646 -> 0.1721 interceptionWon: 0.4962 -> 0.4891 fouls: 1.1313 -> 1.125
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Juan Iglesias', 'matchup_zone': 'Left Wing vs Right Back — Flank Battle', 'duel_win_rate': 0.671, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.2805, 'skill_edge': 0.1788, 'height_edge': 0.012, 'attacker_skill': 0.8995, 'defender_skill': 0.5021, 'attacker_role': 'LW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Juan Iglesias', 'market_value': 5800000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.5021}], 'xg_override': 0.236}	totalShots: 1.9298 -> 2.4711 aerialWon: 0.6719 -> 1.2 shotOffTarget: 1.1903 -> 1.5242 onTargetScoringAttempt: 0.7366 -> 0.9432 expectedGoalsOnTarget: 0.2689 -> 0.3443 expectedGoals: 0.1843 -> 0.236 goals: 0.1721 -> 0.2204 bigChanceMissed: 0.1552 -> 0.1987
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.236 -> 0.4407 expected_assists: 0.1276 -> 0.3074
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 90.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.3564` `goal_rate=0.2204` `scorer_weight=0.4167`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.2646` `assist_rate=0.1265` `provider_weight=0.3055`	-

Arda Güler M 70 0 0.38 0.12 217% 2 1.2 7.1 67.4 25 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0126, quality 0.892)	`market_value_eur=97000000` `quality_factor=0.8922` `boost_factor=1.0126` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0075)	`blend_weight=0.0075` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.778` `prior_reliability=0.1067` `competition=La Liga`	wasFouled: 1.8538 -> 1.8472 fouls: 0.8281 -> 0.8219 wonTackle: 0.6663 -> 0.6723
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Gabriel Suazo', 'matchup_zone': 'Right Wing vs Left Back — Flank Battle', 'duel_win_rate': 0.634, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.2262, 'skill_edge': 0.1619, 'height_edge': -0.008, 'attacker_skill': 0.8922, 'defender_skill': 0.5325, 'attacker_role': 'RW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Gabriel Suazo', 'market_value': 7600000, 'role': 'LB', 'price_index': 0.7, 'adjusted_quality': 0.5325}], 'xg_override': 0.205}	aerialWon: 0.2056 -> 0.9 totalShots: 1.7232 -> 2.113 shotOffTarget: 1.0984 -> 1.3469 onTargetScoringAttempt: 0.6248 -> 0.7661 expectedGoals: 0.1672 -> 0.205 goals: 0.1567 -> 0.1921 bigChanceMissed: 0.129 -> 0.1582 expectedGoalsOnTarget: 0.0427 -> 0.0524
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.205 -> 0.3828 expected_assists: 0.2089 -> 0.5034
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 83.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.3181` `goal_rate=0.1921` `scorer_weight=0.3634`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=1` `assist_probability=0.3955` `assist_rate=0.2007` `provider_weight=0.4874`	-

Aurélien Tchouaméni M 80 0 0.11 0.07 70% 1 1.3 7.0 63.1 24 XGBoost M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0132, quality 0.875)	`market_value_eur=73000000` `quality_factor=0.8747` `boost_factor=1.0132` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Youssouf Fofana', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.551, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.192, 'skill_edge': 0.1186, 'height_edge': 0.012, 'attacker_skill': 0.8865, 'defender_skill': 0.6229, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Youssouf Fofana', 'market_value': 16300000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.6229}], 'xg_override': 0.0601}	aerialWon: 1.3959 -> 1.2 totalShots: 1.0191 -> 1.2148 shotOffTarget: 0.8209 -> 0.9785 onTargetScoringAttempt: 0.1982 -> 0.2363 goals: 0.0621 -> 0.074 expectedGoals: 0.0504 -> 0.0601 bigChanceMissed: 0.0485 -> 0.0578 expectedGoalsOnTarget: 0.0417 -> 0.0497
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0601 -> 0.1122 expected_assists: 0.0472 -> 0.1137
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 84.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1061` `goal_rate=0.074` `scorer_weight=0.1264`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.1075` `assist_rate=0.0525` `provider_weight=0.1199`	-

Federico Valverde M 83 0 0.14 0.08 88% 1 1.3 7.1 68.6 26 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0125, quality 0.887)	`market_value_eur=87000000` `quality_factor=0.8873` `boost_factor=1.0125` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0132)	`blend_weight=0.0132` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=0.927` `prior_reliability=0.1867` `competition=La Liga`	wonTackle: 0.7228 -> 0.7351 wasFouled: 0.6621 -> 0.6534
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Youssouf Fofana, Lucien Agoumé', 'matchup_zone': 'Attacking Mid vs Double Pivot — Tight Space Battle', 'duel_win_rate': 0.338, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.1447, 'skill_edge': 0.1145, 'height_edge': -0.016, 'attacker_skill': 0.8898, 'defender_skill': 0.6355, 'attacker_role': 'CAM', 'attacker_mv_price_index': 0.95, 'defender_adjusted_parts': [{'player': 'Youssouf Fofana', 'market_value': 16300000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.6229}, {'player': 'Lucien Agoumé', 'market_value': 18500000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.648}], 'xg_override': 0.0776}	aerialWon: 0.9303 -> 1.1 totalShots: 0.9324 -> 1.0673 shotOffTarget: 0.6469 -> 0.7405 onTargetScoringAttempt: 0.2855 -> 0.3268 goals: 0.0748 -> 0.0856 expectedGoalsOnTarget: 0.0705 -> 0.0807 expectedGoals: 0.0678 -> 0.0776
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0776 -> 0.1449 expected_assists: 0.1008 -> 0.2429
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 87.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1349` `goal_rate=0.0856` `scorer_weight=0.154`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.2157` `assist_rate=0.0922` `provider_weight=0.2291`	-

Marc Cucurella D 88 0 0.05 0.07 ok 0 0.3 6.8 58.2 26 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0112, quality 0.857)	`market_value_eur=62000000` `quality_factor=0.8569` `boost_factor=1.0112` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0083)	`blend_weight=0.0083` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.9801` `prior_reliability=0.12` `competition=La Liga`	totalCross: 1.5097 -> 1.4972 wasFouled: 0.7827 -> 0.7762 fouls: 0.7316 -> 0.7255
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Chidera Ejuke', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0273 -> 0.051 expected_assists: 0.037 -> 0.0891
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 14 by projected rating, fifa_overall 85.0)	`match_squad_rank=14` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0497` `goal_rate=0.0` `scorer_weight=0.0408`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.0852` `assist_rate=0.0659` `provider_weight=0.1114`	-

Dean Huijsen D 82 0 0.17 0.06 203% 1 0.5 7.0 60.0 25 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0138, quality 0.864)	`market_value_eur=66000000` `quality_factor=0.8644` `boost_factor=1.0138` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0167)	`blend_weight=0.0167` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.9133` `prior_reliability=0.2` `competition=La Liga`	fouls: 0.9422 -> 0.9366 totalShots: 0.5876 -> 0.593 onTargetScoringAttempt: 0.3172 -> 0.3119
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Robbie Ure', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0895 -> 0.1671 expected_assists: 0.0363 -> 0.0875
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 81.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1539` `goal_rate=0.0285` `scorer_weight=0.1337`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.0838` `assist_rate=0.0173` `provider_weight=0.07`	-

Ibrahima Konaté D 89 0 0.08 0.03 218% 1 0.3 6.8 61.3 29 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0095, quality 0.809)	`market_value_eur=46000000` `quality_factor=0.8089` `boost_factor=1.0095` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0090)	`blend_weight=0.009` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.9894` `prior_reliability=0.1333` `competition=La Liga`	wonTackle: 0.7218 -> 0.7331 fouls: 1.1232 -> 1.1175
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Robbie Ure', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0446 -> 0.0833 expected_assists: 0.0124 -> 0.0299
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 84.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0799` `goal_rate=0.0059` `scorer_weight=0.0666`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.0295` `assist_rate=0.0` `provider_weight=0.024`	-

Denzel Dumfries M 84 0 0.18 0.04 359% 1 0.7 6.7 43.4 18 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0226, quality 0.686)	`market_value_eur=26000000` `quality_factor=0.6858` `boost_factor=1.0226` `reference_quality=0.505` `history_matches=7` `history_shrink=0.4167`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0343)	`blend_weight=0.0343` `history_matches=7` `season_overlap_discount=1.0` `per90_to_match_factor=0.9359` `prior_reliability=0.1267` `competition=La Liga`	totalShots: 0.8275 -> 0.8845 fouls: 0.3132 -> 0.3363 wonTackle: 1.3871 -> 1.4071 interceptionWon: 0.5731 -> 0.5872 totalCross: 0.7723 -> 0.7796 onTargetScoringAttempt: 0.3026 -> 0.3092
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Youssouf Fofana', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.477, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.0832, 'skill_edge': 0.0446, 'height_edge': 0.012, 'attacker_skill': 0.7219, 'defender_skill': 0.6229, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Youssouf Fofana', 'market_value': 16300000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.6229}], 'xg_override': 0.0967}	aerialWon: 1.0616 -> 1.2 totalShots: 0.8845 -> 0.9581 shotOffTarget: 0.5249 -> 0.5686 onTargetScoringAttempt: 0.3092 -> 0.3349 bigChanceMissed: 0.2344 -> 0.2539 expectedGoals: 0.0893 -> 0.0967 goals: 0.0739 -> 0.08 expectedGoalsOnTarget: 0.0652 -> 0.0706
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0967 -> 0.1806 expected_assists: 0.0409 -> 0.0985
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 17 by projected rating, fifa_overall 83.0)	`match_squad_rank=17` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1652` `goal_rate=0.08` `scorer_weight=0.16`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.0938` `assist_rate=0.053` `provider_weight=0.1116`	-

Thibaut Courtois G 90 0 0.00 - n/a 0.0 7.1 63.7 24 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0023, quality 0.574)	`market_value_eur=14500000` `quality_factor=0.5737` `boost_factor=1.0023` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0144)	`blend_weight=0.0144` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2` `competition=La Liga`	saves: 2.8736 -> 2.8562
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Robbie Ure, Félix Correia', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.4696}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 3 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=3`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 3.57 (squad xG sum 1.9113, xG scale 1.8673, xA scale 2.4094, team xA 2.32 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=3.569` `squad_xg_sum=1.9113` `squad_xa_sum=0.9628` `team_expected_assists=2.3199` `team_xa_to_xg_ratio=0.65`	shots_faced: 14.0 -> 12.3 expected_saves: 3.43 -> 3.0 high_pressure_saves: 1.18 -> 1.04
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 90.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 3 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=3` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=3.3733`	-
`integer_assist_projection`	Team scoreline allocation: 2 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=2` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=2.4511`	-

Team totals 3 3.57 19 7.0

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Andriy Lunin	G	4.8m	12.7%	Market-value substitute prior
Trent Alexander-Arnold	D	12.9m	44.3%	Market-value substitute prior
Álvaro Carreras	D	15.1m	24.9%	Market-value substitute prior
Eduardo Camavinga	M	17.2m	59.4%	Market-value substitute prior
Franco Mastantuono	M	23.4m	56.4%	Market-value substitute prior
Vinícius Júnior	F	27.1m	37.1%	Market-value substitute prior

Sevilla — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Robbie Ure F 78 1 0.26 0.10 158% 2 1.1 6.9 47.0 10 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9976, quality 0.484)	`market_value_eur=6800000` `quality_factor=0.4842` `boost_factor=0.9976` `reference_quality=0.505` `history_matches=8` `history_shrink=0.3846`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0417)	`blend_weight=0.0417` `history_matches=8` `season_overlap_discount=1.0` `per90_to_match_factor=0.8656` `prior_reliability=0.1667` `competition=La Liga`	totalShots: 2.2672 -> 2.2304 onTargetScoringAttempt: 0.7505 -> 0.7192 wasFouled: 2.3816 -> 2.3545 totalOffside: 0.6261 -> 0.6 goals: 0.2737 -> 0.2623 totalCross: 0.1943 -> 0.1862
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Dean Huijsen, Ibrahima Konaté', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.7084, 'skill_edge': -0.1744, 'height_edge': -0.024, 'attacker_skill': 0.4842, 'defender_skill': 0.8717, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Dean Huijsen', 'market_value': 66000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8879}, {'player': 'Ibrahima Konaté', 'market_value': 46000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8555}], 'xg_override': 0.1786}	totalShots: 2.2359 -> 1.5839 shotOffTarget: 1.5167 -> 1.0744 onTargetScoringAttempt: 0.7192 -> 0.5095 bigChanceMissed: 0.4342 -> 0.3076 goals: 0.2623 -> 0.1858 expectedGoalsOnTarget: 0.2607 -> 0.1847 expectedGoals: 0.2521 -> 0.1786
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.1786 -> 0.2572 expected_assists: 0.0874 -> 0.1203
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 71.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.2268` `goal_rate=0.1858` `scorer_weight=0.2653`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1133` `assist_rate=0.4305` `provider_weight=0.1504`	-

Félix Correia F 75 0 0.22 0.10 128% 1 0.5 6.8 36.7 13 XGBoost F_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0023, quality 0.519)	`market_value_eur=9700000` `quality_factor=0.5189` `boost_factor=1.0023` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Dean Huijsen, Ibrahima Konaté', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.6785, 'skill_edge': -0.1587, 'height_edge': -0.06, 'attacker_skill': 0.5189, 'defender_skill': 0.8717, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Dean Huijsen', 'market_value': 66000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8879}, {'player': 'Ibrahima Konaté', 'market_value': 46000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8555}], 'xg_override': 0.1556}	totalShots: 1.5124 -> 1.0262 shotOffTarget: 0.8241 -> 0.5592 onTargetScoringAttempt: 0.6883 -> 0.467 bigChanceMissed: 0.3896 -> 0.2643 expectedGoalsOnTarget: 0.2885 -> 0.1957 expectedGoals: 0.2293 -> 0.1556 goals: 0.2234 -> 0.1516
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.1556 -> 0.2241 expected_assists: 0.0942 -> 0.1297
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 75.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.2008` `goal_rate=0.1516` `scorer_weight=0.2197`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1216` `assist_rate=0.2673` `provider_weight=0.1621`	-

Youssouf Fofana M 84 0 0.14 0.09 49% 1 1.1 6.9 52.7 7 XGBoost M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0132, quality 0.593)	`market_value_eur=16300000` `quality_factor=0.5932` `boost_factor=1.0132` `reference_quality=0.505` `history_matches=5` `history_shrink=0.5`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Aurélien Tchouaméni', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.289, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.808, 'skill_edge': -0.1186, 'height_edge': -0.012, 'attacker_skill': 0.6229, 'defender_skill': 0.8865, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Aurélien Tchouaméni', 'market_value': 73000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8865}], 'xg_override': 0.0981}	totalShots: 1.5808 -> 1.2773 shotOffTarget: 1.0591 -> 0.8558 aerialWon: 1.0537 -> 1.2 onTargetScoringAttempt: 0.5217 -> 0.4215 bigChanceMissed: 0.3938 -> 0.3182 expectedGoalsOnTarget: 0.2809 -> 0.227 goals: 0.1455 -> 0.1176 expectedGoals: 0.1214 -> 0.0981
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0981 -> 0.1413 expected_assists: 0.1029 -> 0.1417
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 80.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1318` `goal_rate=0.1176` `scorer_weight=0.1599`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1321` `assist_rate=0.1197` `provider_weight=0.1573`	-

Chidera Ejuke F 71 0 0.11 0.03 220% 1 0.5 6.7 49.3 29 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9977, quality 0.446)	`market_value_eur=3700000` `quality_factor=0.4461` `boost_factor=0.9977` `reference_quality=0.505` `history_matches=33` `history_shrink=0.1316`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0057)	`blend_weight=0.0057` `history_matches=33` `season_overlap_discount=1.0` `per90_to_match_factor=0.7914` `prior_reliability=0.0667` `competition=La Liga`	interceptionWon: 0.87 -> 0.865
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Dean Huijsen, Ibrahima Konaté', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v2', 'xg_multiplier': 0.6303, 'skill_edge': -0.1915, 'height_edge': -0.06, 'attacker_skill': 0.4461, 'defender_skill': 0.8717, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Dean Huijsen', 'market_value': 66000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8879}, {'player': 'Ibrahima Konaté', 'market_value': 46000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.8555}], 'xg_override': 0.0731}	totalShots: 1.1324 -> 0.7138 shotOffTarget: 0.7507 -> 0.4732 onTargetScoringAttempt: 0.3817 -> 0.2406 bigChanceMissed: 0.1591 -> 0.1003 goals: 0.1262 -> 0.0795 expectedGoals: 0.116 -> 0.0731 expectedGoalsOnTarget: 0.0169 -> 0.0107
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0731 -> 0.1053 expected_assists: 0.0691 -> 0.0951
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 17 by projected rating, fifa_overall 76.0)	`match_squad_rank=17` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0999` `goal_rate=0.0795` `scorer_weight=0.1108`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0907` `assist_rate=0.0` `provider_weight=0.0761`	-

Lucien Agoumé M 88 0 0.05 0.02 148% 1 0.5 6.9 63.6 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0035, quality 0.616)	`market_value_eur=18500000` `quality_factor=0.6161` `boost_factor=1.0035` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0138)	`blend_weight=0.0138` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.9771` `prior_reliability=0.2` `competition=La Liga`	wonTackle: 1.7692 -> 1.7898 wasFouled: 0.58 -> 0.5945 totalCross: 0.2626 -> 0.277 interceptionWon: 1.3197 -> 1.3285
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Aurélien Tchouaméni', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.301, 'pressing_faced': 'ULTRA_HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8246, 'skill_edge': -0.1073, 'height_edge': -0.012, 'attacker_skill': 0.648, 'defender_skill': 0.8865, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Aurélien Tchouaméni', 'market_value': 73000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8865}], 'xg_override': 0.035}	aerialWon: 0.5964 -> 1.2 totalShots: 0.674 -> 0.5558 shotOffTarget: 0.4969 -> 0.4097 onTargetScoringAttempt: 0.1748 -> 0.1441 goals: 0.0459 -> 0.0378 expectedGoals: 0.0424 -> 0.035
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.035 -> 0.0504 expected_assists: 0.0443 -> 0.061
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 76.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0492` `goal_rate=0.0378` `scorer_weight=0.0521`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0592` `assist_rate=0.058` `provider_weight=0.0699`	-

Arouna Sangante D 76 0 0.00 - n/a 0 0.0 6.7 31.4 8 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9983, quality 0.495)	`market_value_eur=7700000` `quality_factor=0.4951` `boost_factor=0.9983` `reference_quality=0.505` `history_matches=4` `history_shrink=0.5556`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0409)	`blend_weight=0.0409` `history_matches=4` `season_overlap_discount=1.0` `per90_to_match_factor=0.8448` `prior_reliability=0.1133` `competition=La Liga`	interceptionWon: 0.954 -> 0.976 totalShots: 0.4219 -> 0.4046 totalCross: 0.1924 -> 0.2049 wonTackle: 0.7847 -> 0.7729 fouls: 1.2246 -> 1.2355 wasFouled: 0.2662 -> 0.2553
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kylian Mbappé', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 16 by projected rating, fifa_overall 74.0)	`match_squad_rank=16` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `goal_rate=0.0` `scorer_weight=0.001`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0033` `assist_rate=0.0` `provider_weight=0.0031`	-

Gabriel Suazo D 88 0 0.00 0.01 ok 0 0.1 6.8 57.3 28 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9996, quality 0.494)	`market_value_eur=7600000` `quality_factor=0.4939` `boost_factor=0.9996` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0155)	`blend_weight=0.0155` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.9741` `prior_reliability=0.2` `competition=La Liga`	fouls: 1.8297 -> 1.8466 totalCross: 2.7379 -> 2.7508 wonTackle: 1.6567 -> 1.6512
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Arda Güler', 'matchup_zone': 'Right Wing vs Left Back — Flank Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_assists: 0.1715 -> 0.2362
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 74.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0049` `goal_rate=0.0` `scorer_weight=0.0046`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.2104` `assist_rate=0.0022` `provider_weight=0.189`	-

Kike Salas D 90 0 0.07 0.06 ok 1 0.8 6.9 64.9 28 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0032, quality 0.584)	`market_value_eur=15400000` `quality_factor=0.5835` `boost_factor=1.0032` `reference_quality=0.505` `history_matches=32` `history_shrink=0.1351`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0117)	`blend_weight=0.0117` `history_matches=32` `season_overlap_discount=1.0` `per90_to_match_factor=0.9983` `prior_reliability=0.1333` `competition=La Liga`	wasFouled: 0.7602 -> 0.7688 totalShots: 0.9649 -> 0.9715
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kylian Mbappé', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0489 -> 0.0704 expected_assists: 0.0394 -> 0.0542
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 76.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.068` `goal_rate=0.0466` `scorer_weight=0.0688`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0528` `assist_rate=0.0` `provider_weight=0.0434`	-

Andrés Castrin D 89 0 0.04 0.01 156% 1 0.3 6.8 46.2 27 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0001, quality 0.507)	`market_value_eur=8700000` `quality_factor=0.5071` `boost_factor=1.0001` `reference_quality=0.505` `history_matches=26` `history_shrink=0.1613`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kylian Mbappé', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.026 -> 0.0374
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 73.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0367` `goal_rate=0.0504` `scorer_weight=0.0467`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0134` `assist_rate=0.0` `provider_weight=0.0113`	-

Juan Iglesias D 90 0 0.07 0.03 112% 1 0.7 6.8 65.3 31 XGBoost D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9990, quality 0.472)	`market_value_eur=5800000` `quality_factor=0.472` `boost_factor=0.999` `reference_quality=0.505` `history_matches=45` `history_shrink=0.1`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Jude Bellingham', 'matchup_zone': 'Left Wing vs Right Back — Flank Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	expected_goals: 0.0472 -> 0.068 expected_assists: 0.0678 -> 0.0933
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 76.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0657` `goal_rate=0.0966` `scorer_weight=0.085`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0891` `assist_rate=0.0068` `provider_weight=0.0746`	-

Odysseas Vlachodimos G 90 0 0.00 - n/a 0.0 7.0 62.6 31 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9979, quality 0.440)	`market_value_eur=3200000` `quality_factor=0.4399` `boost_factor=0.9979` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0141)	`blend_weight=0.0141` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.2` `competition=La Liga`	saves: 2.934 -> 2.9396
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Kylian Mbappé, Jude Bellingham', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 3.2715}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.30 (squad xG sum 0.9026, xG scale 1.4403, xA scale 1.3766, team xA 1.04 = 0.80 x lambda)	`scoreline_reset=true` `team_lambda=1.3` `squad_xg_sum=0.9026` `squad_xa_sum=0.7555` `team_expected_assists=1.04` `team_xa_to_xg_ratio=0.8`	shots_faced: 31.1 -> 33.9 expected_saves: 7.61 -> 8.3 high_pressure_saves: 2.62 -> 2.86
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 79.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.4226`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=1.0302`	-

Team totals 1 1.30 13 6.8

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Fran González	G	4.4m	10.9%	Market-value substitute prior
José Ángel Carmona	D	14.0m	31.6%	Market-value substitute prior
Marcão	D	13.6m	29.9%	Market-value substitute prior
Oso	M	22.3m	51.5%	Market-value substitute prior
Djibril Sow	M	22.2m	50.8%	Market-value substitute prior
Lucas Stassin	F	25.9m	58.6%	Market-value substitute prior

Antagonist / tactical summary

Home midfield dominates with ~52% possession expected (Jude Bellingham vs Lucien Agoumé). Total antagonist xG — Home: 1.63, Away: 0.54. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Midfield — possession 52.3% / 47.7%, dominance ratio 0.61, key duel: Jude Bellingham vs Lucien Agoumé

Jude Bellingham xG 0.24Arda Güler xG 0.20Federico Valverde xG 0.08Robbie Ure xG 0.18Félix Correia xG 0.16Chidera Ejuke xG 0.07

Summary

Antagonist matchup analysis for Real Madrid vs Sevilla. Most likely scoreline: 3-1. Lambdas derived from 7 home and 5 away individual player duels. Head-to-head: Real Madrid 9W-1D-0L vs Sevilla over 10 meetings.

Tactical parity between 4-2-3-1 and 4-2-3-1. Match outcome driven primarily by individual player quality. Home midfield dominates with ~52% possession expected (Jude Bellingham vs Lucien Agoumé). Total antagonist xG — Home: 1.63, Away: 0.54. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.

Bournemouth vs Sunderland

Premier League · 20261018 · 4-2-3-1 vs 4-2-3-1

Predicted 1-1Winner: DrawWinning H 1-0 (13.2%)Winning A 0-1 (10.1%)Confidence 0.71λ home/away 1.28 / 0.99Most-likely P 13.2%

Actual vs predicted: result not available yet (upcoming fixture).

XGBoost Reasoning & FBref Dataset Tracing

Antagonist + XGBoost match analysis for Bournemouth vs Sunderland. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-2-3-1 vs 4-2-3-1 (duel lambdas: Home 1.279, Away 0.986). Final calibrated Poisson lambdas: Home 1.279, Away 0.986 (most likely score 1-1).

Home Starters 11/11 XGBoost (9 FBref) Away Starters 10/11 XGBoost (9 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.28 / 0.99 Final λ (H/A) 1.28 / 0.99

FBref Team Priors: Bournemouth (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 0.69) vs Sunderland (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 0.69) | Blend Weight: 0.192

Modeling Pipeline Stages (5 layers)

1. XGBoost player regressions for volume & event rates with market-value quality scaling
2. Multi-source prior blending (FBref 2026-2027 tables + Sofascore season history)
3. Antagonist formation duel matrix resolving 1v1 and zonal matchups
4. Team-level prior calibration (FBref possession and attack/defense efficiencies)
5. Bivariate Poisson matrix with Dixon-Coles low-score correction

Standing & form context

Metric	Bournemouth	Sunderland
League position	18.00	17.00
Points per game	0.92	1.00
Attack index	0.93	0.89
Defense (concede) index	1.12	1.16
Form multiplier	0.99	1.00
Standing λ multiplier	1.02	1.01

Lambda decomposition (how the goal rates were built)

Stage	Bournemouth λ	Sunderland λ
FBref Squad Prior E[λ] (weight 0.192)	0.690	0.690
Pre-competitive-floor	1.279	0.986
Pre-calibration (Micro-duels)	1.279	0.986
Final λ	1.279	0.986

Scoreline distribution & markets

Top 5 scorelines	P
1-1	13.2%
1-0	13.2%
0-0	10.5%
0-1	10.1%
2-0	8.5%

Home win	43.0%
Draw	28.5%
Away win	28.5%
Over 2.5	39.5%
BTTS yes	45.4%

Pregame fair line — favorite Bournemouth (43.0%), fair total 2.27, fair spread 0.29

Projected players vs season baseline

Proj = projected for this match · Season = this season's per-game average from match data · deviation badge compares projected xG to the player's season xG/game.

Bournemouth — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Evanilson F 84 0 0.25 0.25 ok 2 1.6 6.8 55.5 30 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0075, quality 0.739)	`market_value_eur=33000000` `quality_factor=0.7389` `boost_factor=1.0075` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0078)	`blend_weight=0.0078` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.9303` `prior_reliability=0.1133` `competition=Premier League`	fouls: 1.3208 -> 1.3319 wasFouled: 1.7877 -> 1.7951 interceptionWon: 0.7612 -> 0.7681 totalOffside: 0.8307 -> 0.8242 goalAssist: 0.2772 -> 0.2836 onTargetScoringAttempt: 0.7203 -> 0.7147
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kevin Danso, Trai Hume', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.207, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 1.0101, 'skill_edge': 0.0229, 'height_edge': -0.016, 'attacker_skill': 0.7389, 'defender_skill': 0.6881, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Kevin Danso', 'market_value': 19400000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6847}, {'player': 'Trai Hume', 'market_value': 20000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6914}], 'xg_override': 0.2975}	totalShots: 1.9462 -> 1.9659 shotOffTarget: 1.2283 -> 1.2407 onTargetScoringAttempt: 0.7147 -> 0.7219
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.2975 -> 0.2532
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 79.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.2237` `goal_rate=0.2691` `scorer_weight=0.2325`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0655` `assist_rate=0.2836` `provider_weight=0.0846`	-

Ben Gannon-Doak F 80 0 0.08 0.01 616% 1 0.2 6.8 18.5 25 FBref Prior F_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0058, quality 0.575)	`market_value_eur=14600000` `quality_factor=0.5748` `boost_factor=1.0058` `reference_quality=0.505` `history_matches=13` `history_shrink=0.2778`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kevin Danso, Trai Hume', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.8604, 'skill_edge': -0.051, 'height_edge': -0.044, 'attacker_skill': 0.5748, 'defender_skill': 0.6881, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Kevin Danso', 'market_value': 19400000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6847}, {'player': 'Trai Hume', 'market_value': 20000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6914}], 'xg_override': 0.0976}	totalShots: 1.5117 -> 1.3007 shotOffTarget: 1.1471 -> 0.987 onTargetScoringAttempt: 0.3646 -> 0.3137 expectedGoals: 0.1134 -> 0.0976 goals: 0.0991 -> 0.0853 bigChanceMissed: 0.0784 -> 0.0675
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0976 -> 0.0831
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 12 by projected rating, fifa_overall 71.0)	`match_squad_rank=12` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0797` `goal_rate=0.0853` `scorer_weight=0.0762`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0725` `assist_rate=0.0288` `provider_weight=0.0602`	-

Marcus Tavernier F 83 0 0.16 0.23 ok 2 1.6 7.1 51.6 37 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0063, quality 0.694)	`market_value_eur=27000000` `quality_factor=0.6941` `boost_factor=1.0063` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0096)	`blend_weight=0.0096` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=0.9254` `prior_reliability=0.1333` `competition=Premier League`	wasFouled: 1.0397 -> 1.0475
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kevin Danso, Trai Hume', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 0.9629, 'skill_edge': 0.0027, 'height_edge': -0.028, 'attacker_skill': 0.6941, 'defender_skill': 0.6881, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Kevin Danso', 'market_value': 19400000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6847}, {'player': 'Trai Hume', 'market_value': 20000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6914}], 'xg_override': 0.1911}	totalShots: 1.9718 -> 1.8986 shotOffTarget: 1.0878 -> 1.0474 onTargetScoringAttempt: 0.8811 -> 0.8484 expectedGoalsOnTarget: 0.3468 -> 0.3339 goals: 0.3044 -> 0.2931 expectedGoals: 0.1985 -> 0.1911
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.1911 -> 0.1626
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 79.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1501` `goal_rate=0.2931` `scorer_weight=0.2032`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.1452` `assist_rate=0.024` `provider_weight=0.1255`	-

Rayan F 85 1 0.28 0.14 97% 3 1.5 7.0 49.8 37 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0198, quality 0.848)	`market_value_eur=58000000` `quality_factor=0.8478` `boost_factor=1.0198` `reference_quality=0.505` `history_matches=21` `history_shrink=0.1923`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0167)	`blend_weight=0.0167` `history_matches=21` `season_overlap_discount=1.0` `per90_to_match_factor=0.9433` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 2.1999 -> 2.1632 wasFouled: 0.9051 -> 0.89 wonTackle: 0.7021 -> 0.6904 onTargetScoringAttempt: 0.8723 -> 0.8817 totalShots: 2.3856 -> 2.3935 fouls: 0.3359 -> 0.3303
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Kevin Danso, Trai Hume', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.272, 'pressing_faced': 'MEDIUM', 'numerical_situation': '1v2', 'xg_multiplier': 1.1057, 'skill_edge': 0.0719, 'height_edge': 0.0, 'attacker_skill': 0.8478, 'defender_skill': 0.6881, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Kevin Danso', 'market_value': 19400000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6847}, {'player': 'Trai Hume', 'market_value': 20000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.6914}], 'xg_override': 0.3234}	totalShots: 2.395 -> 2.6482 shotOffTarget: 1.5133 -> 1.6733 onTargetScoringAttempt: 0.8817 -> 0.9749 bigChanceMissed: 0.4434 -> 0.4903 expectedGoals: 0.2925 -> 0.3234 goals: 0.2606 -> 0.2881 expectedGoalsOnTarget: 0.1502 -> 0.1661
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.3234 -> 0.2752
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 3 by projected rating, fifa_overall 79.0)	`match_squad_rank=3` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.2406` `goal_rate=0.2881` `scorer_weight=0.2492`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0818` `assist_rate=0.031` `provider_weight=0.0682`	-

Ryan Christie M 70 0 0.06 0.05 ok 1 0.8 6.8 51.0 25 FBref Prior M_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0000, quality 0.506)	`market_value_eur=8600000` `quality_factor=0.5059` `boost_factor=1.0` `reference_quality=0.505` `history_matches=31` `history_shrink=0.1389`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Noah Sadiki', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.306, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8321, 'skill_edge': -0.1342, 'height_edge': 0.02, 'attacker_skill': 0.5238, 'defender_skill': 0.8221, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Noah Sadiki', 'market_value': 42000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8221}], 'xg_override': 0.0661}	totalShots: 1.0365 -> 0.8625 shotOffTarget: 0.748 -> 0.6224 aerialWon: 1.0894 -> 1.0 onTargetScoringAttempt: 0.2885 -> 0.2401 bigChanceMissed: 0.2638 -> 0.2195 expectedGoals: 0.0794 -> 0.0661 goals: 0.072 -> 0.0599 expectedGoalsOnTarget: 0.0633 -> 0.0527
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0661 -> 0.0563
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 78.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0547` `goal_rate=0.0599` `scorer_weight=0.0533`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0665` `assist_rate=0.0639` `provider_weight=0.0649`	-

Tyler Adams M 81 0 0.02 0.00 403% 0 0.1 6.8 48.6 22 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0092, quality 0.725)	`market_value_eur=31000000` `quality_factor=0.7249` `boost_factor=1.0092` `reference_quality=0.505` `history_matches=31` `history_shrink=0.1389`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0024)	`blend_weight=0.0024` `history_matches=31` `season_overlap_discount=1.0` `per90_to_match_factor=0.9028` `prior_reliability=0.0267` `competition=Premier League`	interceptionWon: 1.6226 -> 1.6296 fouls: 2.3499 -> 2.3442
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Noah Sadiki', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.401, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.9714, 'skill_edge': -0.0275, 'height_edge': 0.008, 'attacker_skill': 0.761, 'defender_skill': 0.8221, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Noah Sadiki', 'market_value': 42000000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.8221}], 'xg_override': 0.0268}	aerialWon: 0.9405 -> 0.9 totalShots: 0.477 -> 0.4634 shotOffTarget: 0.348 -> 0.338
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 9 by projected rating, fifa_overall 79.0)	`match_squad_rank=9` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0225` `goal_rate=0.0376` `scorer_weight=0.026`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0332` `assist_rate=0.0496` `provider_weight=0.0397`	-

António Silva D 90 0 0.21 0.12 81% 1 0.5 7.0 36.7 11 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0310, quality 0.732)	`market_value_eur=32000000` `quality_factor=0.732` `boost_factor=1.031` `reference_quality=0.505` `history_matches=6` `history_shrink=0.4545`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0394)	`blend_weight=0.0394` `history_matches=6` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	onTargetScoringAttempt: 0.4007 -> 0.3849 totalShots: 0.8266 -> 0.8137 fouls: 0.7414 -> 0.7319 wonTackle: 0.7965 -> 0.8045 interceptionWon: 1.1753 -> 1.1684 wasFouled: 0.1525 -> 0.1465
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Nilson Angulo', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.2513 -> 0.2139
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall 77.0)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1926` `goal_rate=0.0` `scorer_weight=0.1711`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.054` `assist_rate=0.0` `provider_weight=0.0444`	-

James Hill D 90 0 0.01 0.02 46% 0 0.3 6.9 58.7 30 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0085, quality 0.732)	`market_value_eur=32000000` `quality_factor=0.732` `boost_factor=1.0085` `reference_quality=0.505` `history_matches=35` `history_shrink=0.125`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0108)	`blend_weight=0.0108` `history_matches=35` `season_overlap_discount=1.0` `per90_to_match_factor=0.9992` `prior_reliability=0.1333` `competition=Premier League`	fouls: 0.9227 -> 0.9289 totalShots: 0.459 -> 0.454
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Nilson Angulo', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 78.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0132` `goal_rate=0.0` `scorer_weight=0.0106`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0736` `assist_rate=0.02` `provider_weight=0.0611`	-

Adam Smith D 82 0 0.00 0.00 142% 0 0.0 6.6 34.8 30 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9969, quality 0.437)	`market_value_eur=310000` `quality_factor=0.4374` `boost_factor=0.9969` `reference_quality=0.505` `history_matches=28` `history_shrink=0.1515`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0131)	`blend_weight=0.0131` `history_matches=28` `season_overlap_discount=1.0` `per90_to_match_factor=0.9122` `prior_reliability=0.1333` `competition=Premier League`	interceptionWon: 0.7459 -> 0.7541 wonTackle: 0.8546 -> 0.8494
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Nilson Angulo', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 18 by projected rating, fifa_overall 73.0)	`match_squad_rank=18` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0012` `goal_rate=0.0` `scorer_weight=0.0012`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0229` `assist_rate=0.0604` `provider_weight=0.029`	-

Đorđe Petrović G 90 0 0.00 - n/a 0.0 6.9 57.3 30 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0063, quality 0.710)	`market_value_eur=29000000` `quality_factor=0.71` `boost_factor=1.0063` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	saves: 2.9235 -> 2.9374
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Enzo Le Fée, Timur Tuterov', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 0.9855}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 79.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=1.2321`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.7617`	-

Adrien Truffert D 89 0 0.04 0.04 ok 1 0.4 6.8 57.4 30 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0072, quality 0.739)	`market_value_eur=33000000` `quality_factor=0.7389` `boost_factor=1.0072` `reference_quality=0.505` `history_matches=44` `history_shrink=0.102`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=44` `season_overlap_discount=1.0` `per90_to_match_factor=0.9899` `prior_reliability=0.1333` `competition=Premier League`	totalCross: 3.8209 -> 3.809 wonTackle: 1.4094 -> 1.4013 totalShots: 0.7423 -> 0.7357
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Granit Xhaka', 'matchup_zone': 'Left Wing vs Right Back — Flank Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 1.28 (squad xG sum 1.5028, xG scale 0.8511, xA scale 0.9726, team xA 0.83 = 0.65 x lambda)	`scoreline_reset=true` `team_lambda=1.279` `squad_xg_sum=1.5028` `squad_xa_sum=0.8548` `team_expected_assists=0.8313` `team_xa_to_xg_ratio=0.65`	expected_goals: 0.0428 -> 0.0364
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 10 by projected rating, fifa_overall 80.0)	`match_squad_rank=10` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0357` `goal_rate=0.0262` `scorer_weight=0.0299`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1059` `assist_rate=0.131` `provider_weight=0.1227`	-

Team totals 1 1.28 14 6.9

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Michele Di Gregorio	G	4.7m	12.3%	Market-value substitute prior
Bafodé Diakité	D	14.4m	33.9%	Market-value substitute prior
Álex Jiménez	D	14.3m	33.3%	Market-value substitute prior
Alex Scott	M	23.7m	34.3%	Market-value substitute prior
Justin Kluivert	M	22.9m	32.6%	Market-value substitute prior
Eli Junior Kroupi	F	27.0m	63.7%	Market-value substitute prior

Sunderland — projected line vs season baseline (4-2-3-1)

Player Pos Min Proj G Proj xG Season xG/g Proj Sh Season Sh/g Proj Rtg Season Rtg Rec Model / FBref Prior

Reinildo Mandava D 85 0 0.00 0.00 48% 0 0.0 6.7 49.3 23 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 0.9977, quality 0.451)	`market_value_eur=4099999` `quality_factor=0.4511` `boost_factor=0.9977` `reference_quality=0.505` `history_matches=30` `history_shrink=0.1429`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0124)	`blend_weight=0.0124` `history_matches=30` `season_overlap_discount=1.0` `per90_to_match_factor=0.9463` `prior_reliability=0.1333` `competition=Premier League`	fouls: 1.3077 -> 1.3208 interceptionWon: 1.185 -> 1.1938 totalCross: 0.7473 -> 0.7556 totalShots: 0.4281 -> 0.4228
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Rayan', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_assists: 0.0396 -> 0.034
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 13 by projected rating, fifa_overall 80.0)	`match_squad_rank=13` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv_filled` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0015` `goal_rate=0.0` `scorer_weight=0.0014`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0334` `assist_rate=0.0181` `provider_weight=0.0272`	-

Kevin Danso D 89 0 0.06 0.03 85% 1 0.5 6.8 38.4 26 FBref Prior D_service

Step-by-step (9 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0055, quality 0.625)	`market_value_eur=19400000` `quality_factor=0.6251` `boost_factor=1.0055` `reference_quality=0.505` `history_matches=28` `history_shrink=0.1515`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Rayan', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.0679 -> 0.0583
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 6 by projected rating, fifa_overall 79.0)	`match_squad_rank=6` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0566` `goal_rate=0.0` `scorer_weight=0.0466`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0241` `assist_rate=0.0` `provider_weight=0.0195`	-

Omar Alderete D 78 0 0.03 0.02 51% 1 0.4 6.8 42.9 31 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0009, quality 0.529)	`market_value_eur=10600000` `quality_factor=0.5295` `boost_factor=1.0009` `reference_quality=0.505` `history_matches=34` `history_shrink=0.1282`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0017)	`blend_weight=0.0017` `history_matches=34` `season_overlap_discount=1.0` `per90_to_match_factor=0.8679` `prior_reliability=0.02` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Marcus Tavernier', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.0363 -> 0.0312
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 11 by projected rating, fifa_overall 81.0)	`match_squad_rank=11` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0307` `goal_rate=0.0022` `scorer_weight=0.025`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0115` `assist_rate=0.0` `provider_weight=0.0093`	-

Nordi Mukiele D 88 0 0.11 0.07 51% 1 0.9 6.8 51.5 28 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0035, quality 0.602)	`market_value_eur=17100000` `quality_factor=0.6016` `boost_factor=1.0035` `reference_quality=0.505` `history_matches=37` `history_shrink=0.119`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0052)	`blend_weight=0.0052` `history_matches=37` `season_overlap_discount=1.0` `per90_to_match_factor=0.9794` `prior_reliability=0.0667` `competition=Premier League`	totalCross: 3.3363 -> 3.3241 fouls: 1.2518 -> 1.2453 wonTackle: 1.1837 -> 1.1776 totalShots: 1.1193 -> 1.1135
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Ben Gannon-Doak', 'matchup_zone': 'Shape fallback'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.1261 -> 0.1083 expected_assists: 0.0862 -> 0.074
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 8 by projected rating, fifa_overall 81.0)	`match_squad_rank=8` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1026` `goal_rate=0.0715` `scorer_weight=0.0866`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0713` `assist_rate=0.0353` `provider_weight=0.0592`	-

Robin Roefs G 90 0 0.00 - n/a 0.0 7.0 54.2 28 FBref Prior G_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost G_service raw output over 3 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0072, quality 0.725)	`market_value_eur=31000000` `quality_factor=0.7249` `boost_factor=1.0072` `reference_quality=0.505` `history_matches=41` `history_shrink=0.1087`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0094)	`blend_weight=0.0094` `history_matches=41` `season_overlap_discount=1.0` `per90_to_match_factor=1.0` `prior_reliability=0.1333` `competition=Premier League`	saves: 2.882 -> 2.8596
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Rayan, Evanilson', 'matchup_zone': 'Goalkeeper shot-stopping context', 'opponent_lambda': 1.279}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	-
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 2 by projected rating, fifa_overall 81.0)	`match_squad_rank=2` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0` `scorer_weight=0.0` `squad_weight_total=0.9086`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0` `provider_weight=0.0` `squad_weight_total=0.6875`	-

Granit Xhaka M 90 0 0.04 0.02 87% 1 0.6 7.3 60.1 27 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0000, quality 0.505)	`market_value_eur=8500000` `quality_factor=0.5047` `boost_factor=1.0` `reference_quality=0.505` `history_matches=40` `history_shrink=0.1111`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0096)	`blend_weight=0.0096` `history_matches=40` `season_overlap_discount=1.0` `per90_to_match_factor=0.9952` `prior_reliability=0.1333` `competition=Premier League`	wasFouled: 1.6554 -> 1.673 totalShots: 0.9351 -> 0.9261 totalCross: 4.6646 -> 4.6724
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Adrien Truffert', 'matchup_zone': 'Left Wing vs Right Back — Flank Battle', 'duel_win_rate': 0.377, 'pressing_faced': 'HIGH', 'numerical_situation': '1v1', 'xg_multiplier': 0.8486, 'skill_edge': -0.139, 'height_edge': 0.036, 'attacker_skill': 0.5047, 'defender_skill': 0.8135, 'attacker_role': 'LW', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'Adrien Truffert', 'market_value': 33000000, 'role': 'RB', 'price_index': 0.7, 'adjusted_quality': 0.8135}], 'xg_override': 0.051}	aerialWon: 0.7882 -> 1.2 totalShots: 0.9324 -> 0.7912 shotOffTarget: 0.6672 -> 0.5662 onTargetScoringAttempt: 0.2652 -> 0.225 expectedGoalsOnTarget: 0.0706 -> 0.0599 goals: 0.0665 -> 0.0564 expectedGoals: 0.0601 -> 0.051
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.051 -> 0.0438 expected_assists: 0.1935 -> 0.1661
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 1 by projected rating, fifa_overall 85.0)	`match_squad_rank=1` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0429` `goal_rate=0.0564` `scorer_weight=0.0461`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=1` `assist_probability=0.153` `assist_rate=0.1285` `provider_weight=0.1329`	-

Noah Sadiki M 84 0 0.08 0.06 35% 1 0.7 6.8 55.6 28 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0097, quality 0.791)	`market_value_eur=42000000` `quality_factor=0.7909` `boost_factor=1.0097` `reference_quality=0.505` `history_matches=39` `history_shrink=0.1136`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0098)	`blend_weight=0.0098` `history_matches=39` `season_overlap_discount=1.0` `per90_to_match_factor=0.9373` `prior_reliability=0.1333` `competition=Premier League`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ryan Christie', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.534, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.1679, 'skill_edge': 0.1342, 'height_edge': -0.02, 'attacker_skill': 0.8221, 'defender_skill': 0.5238, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Ryan Christie', 'market_value': 8600000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.5238}], 'xg_override': 0.091}	aerialWon: 0.7117 -> 0.9 totalShots: 0.9821 -> 1.147 shotOffTarget: 0.6548 -> 0.7647 onTargetScoringAttempt: 0.3273 -> 0.3823 bigChanceMissed: 0.2108 -> 0.2462 expectedGoals: 0.0779 -> 0.091 goals: 0.0747 -> 0.0872 expectedGoalsOnTarget: 0.0417 -> 0.0487
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.091 -> 0.0781 expected_assists: 0.0899 -> 0.0772
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 7 by projected rating, fifa_overall 80.0)	`match_squad_rank=7` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0751` `goal_rate=0.0872` `scorer_weight=0.076`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0743` `assist_rate=0.0768` `provider_weight=0.07`	-

Enzo Le Fée M 88 1 0.21 0.17 ok 2 1.1 6.9 58.7 30 FBref Prior M_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost M_service raw output over 8 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0063, quality 0.702)	`market_value_eur=28000000` `quality_factor=0.7022` `boost_factor=1.0063` `reference_quality=0.505` `history_matches=42` `history_shrink=0.1064`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0088)	`blend_weight=0.0088` `history_matches=42` `season_overlap_discount=1.0` `per90_to_match_factor=0.9743` `prior_reliability=0.1267` `competition=Premier League`	totalCross: 3.2348 -> 3.2244
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'Ryan Christie', 'matchup_zone': 'CDM vs CDM', 'duel_win_rate': 0.497, 'pressing_faced': 'LOW', 'numerical_situation': '1v1', 'xg_multiplier': 1.1126, 'skill_edge': 0.0966, 'height_edge': -0.02, 'attacker_skill': 0.7385, 'defender_skill': 0.5238, 'attacker_role': 'CDM', 'attacker_mv_price_index': 0.85, 'defender_adjusted_parts': [{'player': 'Ryan Christie', 'market_value': 8600000, 'role': 'CDM', 'price_index': 0.85, 'adjusted_quality': 0.5238}], 'xg_override': 0.2473}	aerialWon: 0.4364 -> 0.9 totalShots: 1.7139 -> 1.9069 shotOffTarget: 1.1094 -> 1.2343 onTargetScoringAttempt: 0.6005 -> 0.6681 expectedGoalsOnTarget: 0.2293 -> 0.2551 expectedGoals: 0.2223 -> 0.2473 bigChanceMissed: 0.2029 -> 0.2257 goals: 0.193 -> 0.2147
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.2473 -> 0.2124 expected_assists: 0.1617 -> 0.1389
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 5 by projected rating, fifa_overall 80.0)	`match_squad_rank=5` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=1` `goal_probability=0.1914` `goal_rate=0.2147` `scorer_weight=0.1891`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.1297` `assist_rate=0.1307` `provider_weight=0.1186`	-

Trai Hume D 86 0 0.04 0.06 ok 1 0.9 6.7 54.8 33 FBref Prior D_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost D_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0039, quality 0.631)	`market_value_eur=20000000` `quality_factor=0.6311` `boost_factor=1.0039` `reference_quality=0.505` `history_matches=43` `history_shrink=0.1042`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0086)	`blend_weight=0.0086` `history_matches=43` `season_overlap_discount=1.0` `per90_to_match_factor=0.9548` `prior_reliability=0.1267` `competition=Premier League`	wasFouled: 0.8761 -> 0.8686 wonTackle: 1.1474 -> 1.1548
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` `duel_context={'opponent_faced': 'Rayan', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle'}`	-
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.0515 -> 0.0442 expected_assists: 0.0903 -> 0.0775
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 17 by projected rating, fifa_overall 78.0)	`match_squad_rank=17` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.0432` `goal_rate=0.0996` `scorer_weight=0.0553`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0746` `assist_rate=0.0013` `provider_weight=0.062`	-

Timur Tuterov F 90 0 0.17 - n/a 2 0.0 6.9 - 8 Prior F_service

Step-by-step (8 steps)

Step	Description	Key params	Changes
`market_value_prior`	Market-value baseline for position F (quality 0.44)	`market_value_eur=280000` `quality_factor=0.4374`	-
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'James Hill, Adam Smith', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.18, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 0.8879, 'skill_edge': -0.0803, 'height_edge': 0.004, 'attacker_skill': 0.4374, 'defender_skill': 0.6158, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'James Hill', 'market_value': 32000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.7941}, {'player': 'Adam Smith', 'market_value': 310000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.2007}	aerialWon: 1.4 -> 1.7 totalShots: 2.4 -> 2.131 onTargetScoringAttempt: 1.2 -> 1.0655 shotOffTarget: 0.8 -> 0.7103 expectedGoals: 0.226 -> 0.2007 expectedGoalsOnTarget: 0.171 -> 0.1518 goals: 0.14 -> 0.1243 bigChanceMissed: 0.14 -> 0.1243
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.2007 -> 0.1724 expected_assists: 0.081 -> 0.0696
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 4 by projected rating, fifa_overall n/a)	`match_squad_rank=4` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=missing_player_rankings_csv_row` `ranking_status=missing`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.1584` `goal_rate=0.1243` `scorer_weight=0.1379`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0672` `assist_rate=0.07` `provider_weight=0.0638`	-

Nilson Angulo F 70 0 0.15 0.07 127% 2 0.5 6.7 43.4 30 FBref Prior F_service

Step-by-step (10 steps)

Step	Description	Key params	Changes
`xgboost_raw`	XGBoost F_service raw output over 7 target models	-	-
`market_value_boost`	Market-value quality scaling (boost 1.0095, quality 0.625)	`market_value_eur=19400000` `quality_factor=0.6251` `boost_factor=1.0095` `reference_quality=0.505` `history_matches=14` `history_shrink=0.2632`	-
`fbref_prior_blend`	FBref 2026-27 prior blended (weight 0.0160)	`blend_weight=0.016` `history_matches=14` `season_overlap_discount=1.0` `per90_to_match_factor=0.7763` `prior_reliability=0.0933` `competition=Premier League`	wasFouled: 0.6967 -> 0.721 totalCross: 1.1302 -> 1.1476 goals: 0.1801 -> 0.1861
`final_cap`	Stats clamped to historical constraints and shot-component identity	`stats_clamped=0` `shot_identity_applied=false`	-
`enriched_model_output`	Model output after duel-context enrichment (pre router post-processing)	`duel_context_applied=true` duel_context={'opponent_faced': 'James Hill, Adam Smith', 'matchup_zone': 'Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle', 'duel_win_rate': 0.204, 'pressing_faced': 'LOW', 'numerical_situation': '1v2', 'xg_multiplier': 1.0062, 'skill_edge': 0.0042, 'height_edge': 0.0, 'attacker_skill': 0.6251, 'defender_skill': 0.6158, 'attacker_role': 'ST', 'attacker_mv_price_index': 1.0, 'defender_adjusted_parts': [{'player': 'James Hill', 'market_value': 32000000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.7941}, {'player': 'Adam Smith', 'market_value': 310000, 'role': 'CB', 'price_index': 0.75, 'adjusted_quality': 0.4374}], 'xg_override': 0.1743}	totalShots: 1.3031 -> 1.3112
`scoreline_reset`	Team scoreline allocation: goal/assist counters reset to 0; the team's 1 goals are handed out as whole numbers by integer_goal_projection / integer_assist_projection	`scoreline_reset=true` `team_goals=1`	-
`team_lambda_rescale`	Player expected_goals/expected_assists rescaled to unconditional contributions so squad sum equals team lambda 0.99 (squad xG sum 1.1482, xG scale 0.8588, xA scale 0.8588, team xA 0.78 = 0.79 x lambda)	`scoreline_reset=true` `team_lambda=0.986` `squad_xg_sum=1.1482` `squad_xa_sum=0.9056` `team_expected_assists=0.7777` `team_xa_to_xg_ratio=0.7887`	expected_goals: 0.1743 -> 0.1497 expected_assists: 0.0618 -> 0.0531
`rankings_attached`	Display-only provenance: no published number changes. Global ranking attached (match squad rank 12 by projected rating, fifa_overall 73.0)	`match_squad_rank=12` `squad_role=starter` `squad_rank_basis=projected_rating` `affects_projection=false` `ranking_source=sofascore_plus_fifa_csv` `ranking_status=combined`	-
`integer_goal_projection`	Team scoreline allocation: 1 projected goals distributed by scorer weight (xG x finishing x appearance)	`team_goals=1` `allocation_applied=true` `goals_allocated=0` `goal_probability=0.139` `goal_rate=0.1873` `scorer_weight=0.1584`	-
`integer_assist_projection`	Team scoreline allocation: 1 projected assists distributed by provider weight (xA x creation x appearance; nobody assists his own goal)	`allocation_applied=true` `team_assists=1` `assists_allocated=0` `assist_probability=0.0517` `assist_rate=0.1521` `provider_weight=0.0664`	-

Team totals 1 0.99 14 6.9

Bench Candidates (7 modeled substitutes with empirical playing time)

Player	Pos	Exp Min	App Prob	Substitute Modeling Details
Melker Ellborg	G	4.4m	10.9%	Market-value substitute prior
Daniel Ballard	D	14.6m	23.8%	Market-value substitute prior
Lutsharel Geertruida	D	14.4m	33.7%	Market-value substitute prior
Habib Diarra	M	18.7m	71.8%	Market-value substitute prior
Dayann Methalie	M	23.0m	54.5%	Market-value substitute prior
Brian Brobbey	F	26.9m	36.6%	Market-value substitute prior

Antagonist / tactical summary

Midfield is evenly contested — Tyler Adams vs Noah Sadiki will be the pivotal duel. Total antagonist xG — Home: 1.00, Away: 0.76. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: CDM vs CDM.

Midfield — possession 48.7% / 51.3%, dominance ratio 0.47, key duel: Tyler Adams vs Noah Sadiki

Rayan xG 0.32Evanilson xG 0.30Ben Gannon-Doak xG 0.10Granit Xhaka xG 0.05Nilson Angulo xG 0.17Timur Tuterov xG 0.20

Summary

Antagonist matchup analysis for Bournemouth vs Sunderland. Most likely scoreline: 1-1. Lambdas derived from 6 home and 5 away individual player duels. Head-to-head: Bournemouth 2W-2D-2L vs Sunderland over 6 meetings.

Tactical parity between 4-2-3-1 and 4-2-3-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Tyler Adams vs Noah Sadiki will be the pivotal duel. Total antagonist xG — Home: 1.00, Away: 0.76. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: CDM vs CDM.

Generated 2026-10-18 13:37:42 UTC · Soccer prediction reporting