Soccer predictions: actual data vs predicted data
Slate 20260920 · 9 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 detail
Deportivo vs Real Betis
LaLiga · 20260920 · 3-5-2 vs 3-5-2
Predicted 1-0Winner: DeportivoConfidence 0.59λ home/away 1.62 / 1.13Most-likely P 10.3%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Deportivo vs Real Betis. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 3-5-2 vs 3-5-2 (duel lambdas: Home 0.840, Away 1.129). Final calibrated Poisson lambdas: Home 1.616, Away 1.129 (most likely score 1-0).
Home Starters 10/11 XGBoost (0 FBref) Away Starters 10/11 XGBoost (8 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 0.84 / 1.13 Final λ (H/A) 1.62 / 1.13
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 | Deportivo | Real Betis |
|---|
| League position | 5.00 | 4.00 |
| Points per game | 2.00 | 2.25 |
| Attack index | 1.67 | 1.04 |
| Defense (concede) index | 1.04 | 1.04 |
| Form multiplier | 0.98 | 1.01 |
| Standing λ multiplier | 1.25 | 1.04 |
Lambda decomposition (how the goal rates were built)
| Stage | Deportivo λ | Real Betis λ |
|---|
| Pre-competitive-floor | 0.840 | 1.129 |
| Pre-calibration (Micro-duels) | 0.840 | 1.129 |
| Final λ | 1.616 | 1.129 |
competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Falsecompetitive underdog floor applied Truestage discount 1.0
Scoreline distribution & markets
| Top 5 scorelines | P |
|---|
| 1-0 | 10.3% |
| 1-1 | 11.8% |
| 2-1 | 9.5% |
| 2-0 | 8.4% |
| 0-1 | 7.2% |
| Home win | 48.6% |
| Draw | 24.9% |
| Away win | 26.5% |
| Over 2.5 | 51.7% |
| BTTS yes | 54.3% |
Pregame fair line — favorite Deportivo (48.6%), fair total 2.75, fair spread 0.49
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.
Deportivo — projected line vs season baseline (3-5-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Leo Román | G | 89 | 0 | 0.00 | - | n/a | | 0.0 | 7.0 | 58.4 | 5 | XGBoost G_service |
| José María Giménez | D | 34 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.7 | 40.8 | 5 | XGBoost D_service |
| Giacomo Quagliata | D | 88 | 0 | 0.00 | 0.00 | 100% | 0 | 0.2 | 6.5 | 51.6 | 5 | XGBoost D_service |
| Adrià Altimira | D | 50 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.6 | 50.5 | 4 | XGBoost D_service |
| Marc Casadó | M | 25 | 0 | 0.17 | - | n/a | 2 | 0.0 | 6.8 | 34.0 | 2 | XGBoost M_service |
| Lorenzo Amatucci | M | 90 | 0 | 0.10 | - | n/a | 1 | 0.0 | 6.8 | 56.4 | 5 | XGBoost M_service |
| Mario Soriano | M | 73 | 0 | 0.11 | 0.02 | 549% | 1 | 0.6 | 6.8 | 54.8 | 5 | XGBoost M_service |
| David Mella | M | 23 | 0 | 0.20 | - | n/a | 1 | 0.0 | 6.7 | 26.0 | 5 | XGBoost M_service |
| Angeliño | M | 90 | 0 | 0.28 | - | n/a | 2 | 0.0 | 6.9 | - | 5 | Prior M_service |
| Yeremay Hernández | F | 24 | 0 | 0.25 | 0.02 | 1139% | 2 | 0.2 | 7.0 | 50.0 | 4 | XGBoost F_service |
| Adama Traoré | F | 24 | 1 | 0.28 | - | n/a | 2 | 0.0 | 6.8 | 30.5 | 2 | XGBoost F_service |
| Team totals | 1 | 1.62 | | 11 | | 6.8 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Álvaro Ferllo | G | 4.4m | 10.9% | Market-value substitute prior |
| Lucas Noubi | D | 13.6m | 29.9% | Market-value substitute prior |
| Dani Barcia | D | 13.6m | 29.9% | Market-value substitute prior |
| Luismi Cruz | M | 22.0m | 49.9% | Market-value substitute prior |
| Riki Rodríguez | M | 22.0m | 49.9% | Market-value substitute prior |
| Jonathan Asp Jensen | F | 25.3m | 56.1% | Market-value substitute prior |
Real Betis — projected line vs season baseline (3-5-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Álvaro Valles | G | 89 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.3 | 66.4 | 5 | FBref Prior G_service |
| Natan | D | 95 | 0 | 0.00 | 0.00 | 100% | 0 | 0.2 | 7.1 | 43.6 | 5 | FBref Prior D_service |
| Valentín Gómez | D | 91 | 0 | 0.00 | - | n/a | 1 | 0.0 | 6.7 | 14.6 | 5 | FBref Prior D_service |
| Fran García | D | 76 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.7 | 55.0 | 5 | FBref Prior D_service |
| Abdessamad Ezzalzouli | M | 17 | 0 | 0.10 | - | n/a | 2 | 0.0 | 7.5 | 33.5 | 2 | XGBoost M_service |
| Antony | M | 94 | 0 | 0.21 | 0.18 | ok | 3 | 3.8 | 7.5 | 57.6 | 5 | FBref Prior M_service |
| Nelson Deossa | M | 36 | 0 | 0.07 | 0.04 | 77% | 1 | 1.6 | 6.9 | 54.6 | 5 | FBref Prior M_service |
| Pablo Fornals | M | 73 | 0 | 0.11 | 0.05 | 122% | 1 | 1.2 | 7.0 | 54.6 | 5 | FBref Prior M_service |
| Giovani Lo Celso | M | 90 | 0 | 0.13 | - | n/a | 1 | - | 6.9 | - | 1 | Prior M_service |
| Troy Parrott | F | 34 | 0 | 0.21 | 0.31 | 31% | 2 | 1.0 | 7.1 | 49.5 | 4 | FBref Prior F_service |
| Cucho Hernández | F | 65 | 0 | 0.20 | 0.03 | 534% | 2 | 0.6 | 6.9 | 52.2 | 5 | XGBoost F_service |
| Team totals | 0 | 1.13 | | 13 | | 7.1 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Diego Conde | G | 4.4m | 10.9% | Market-value substitute prior |
| Angel Ortiz | D | 13.6m | 21.8% | Market-value substitute prior |
| Aitor Ruibal | D | 13.6m | 30.1% | Market-value substitute prior |
| Dani Ceballos | M | 22.2m | 50.8% | Market-value substitute prior |
| Álvaro Fidalgo | M | 22.1m | 50.6% | Market-value substitute prior |
| Pablo García | F | 21.4m | 72.4% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Marc Casadó vs Abdessamad Ezzalzouli will be the pivotal duel. Total antagonist xG — Home: 0.38, Away: 0.68. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Central Mid vs Defensive Mid — Pressing/Creation Duel.
Midfield — possession 48.6% / 51.4%, dominance ratio 0.45, key duel: Marc Casadó vs Abdessamad Ezzalzouli
Mario Soriano xG 0.03Angeliño xG 0.08Marc Casadó xG 0.05Nelson Deossa xG 0.06Abdessamad Ezzalzouli xG 0.09Antony xG 0.17
Summary
Antagonist matchup analysis for Deportivo vs Real Betis. Most likely scoreline: 1-0. Lambdas derived from 6 home and 6 away individual player duels.
Tactical parity between 3-5-2 and 3-5-2. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Marc Casadó vs Abdessamad Ezzalzouli will be the pivotal duel. Total antagonist xG — Home: 0.38, Away: 0.68. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Central Mid vs Defensive Mid — Pressing/Creation Duel.
Leeds vs Crystal Palace
Premier League · 20260920 · 4-3-3 vs 3-4-2-1
Predicted 1-0Winner: LeedsConfidence 0.59λ home/away 1.61 / 1.02Most-likely P 11.5%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Leeds vs Crystal Palace. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-3-3 vs 3-4-2-1 (duel lambdas: Home 1.611, Away 0.909). Final calibrated Poisson lambdas: Home 1.611, Away 1.020 (most likely score 1-0).
Home Starters 10/11 XGBoost (10 FBref) Away Starters 9/11 XGBoost (8 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.61 / 0.91 Final λ (H/A) 1.61 / 1.02
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 | Crystal Palace |
|---|
| League position | 9.00 | 14.00 |
| Points per game | 1.67 | 1.00 |
| Attack index | 1.25 | 0.83 |
| Defense (concede) index | 1.67 | 0.83 |
| Form multiplier | 1.03 | 0.96 |
| Standing λ multiplier | 1.36 | 0.86 |
Lambda decomposition (how the goal rates were built)
| Stage | Leeds λ | Crystal Palace λ |
|---|
| Pre-competitive-floor | 1.611 | 0.909 |
| Pre-calibration (Micro-duels) | 1.611 | 0.909 |
| Final λ | 1.611 | 1.020 |
competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Falsecompetitive underdog floor applied Truestage discount 1.0
Scoreline distribution & markets
| Top 5 scorelines | P |
|---|
| 1-0 | 11.5% |
| 1-1 | 11.9% |
| 2-1 | 9.5% |
| 2-0 | 9.3% |
| 0-0 | 7.3% |
| Home win | 51.0% |
| Draw | 25.1% |
| Away win | 23.9% |
| Over 2.5 | 48.9% |
| BTTS yes | 51.3% |
Pregame fair line — favorite Leeds (51.0%), fair total 2.63, fair spread 0.59
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 (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 |
|---|
| James Trafford | G | 95 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.6 | 52.2 | 4 | FBref Prior G_service |
| Tarik Muharemović | D | 95 | 0 | 0.00 | 0.02 | 100% | 0 | 0.2 | 7.2 | 53.0 | 4 | FBref Prior D_service |
| Gabriel Gudmundsson | D | 94 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.9 | 22.7 | 3 | FBref Prior D_service |
| Joe Rodon | D | 64 | 0 | 0.00 | 0.01 | 100% | 1 | 0.3 | 7.0 | 46.3 | 3 | FBref Prior D_service |
| Jaka Bijol | D | 57 | 0 | 0.00 | 0.03 | 100% | 1 | 0.5 | 7.0 | 51.0 | 4 | FBref Prior D_service |
| Ethan Ampadu | M | 95 | 0 | 0.07 | 0.00 | 1948% | 0 | 0.2 | 7.2 | 51.8 | 4 | FBref Prior M_service |
| Harry Wilson | M | 45 | 0 | 0.06 | 0.00 | 2304% | 1 | 0.2 | 7.2 | 50.2 | 4 | FBref Prior M_service |
| Anton Stach | M | 95 | 0 | 0.28 | 0.12 | 126% | 2 | 1.2 | 7.3 | 53.8 | 4 | FBref Prior M_service |
| Noah Okafor | F | 55 | 0 | 0.27 | 0.02 | 1279% | 2 | 0.5 | 7.2 | 50.8 | 4 | FBref Prior F_service |
| Dominic Calvert-Lewin | F | 89 | 0 | 0.35 | 0.40 | ok | 2 | 1.5 | 7.0 | 51.0 | 4 | FBref Prior F_service |
| Jean-Mattéo Bahoya | F | 90 | 1 | 0.45 | - | n/a | 3 | 0.0 | 7.2 | - | 2 | Prior F_service |
| Team totals | 1 | 1.61 | | 13 | | 7.2 | | |
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 |
| James Justin | D | 14.0m | 22.6% | Market-value substitute prior |
| Melvin Bard | D | 13.9m | 31.5% | Market-value substitute prior |
| Jayden Bogle | M | 22.6m | 31.9% | Market-value substitute prior |
| Sean Longstaff | M | 15.1m | 70.3% | Market-value substitute prior |
| Brenden Aaronson | F | 26.2m | 35.2% | 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 |
|---|
| Dean Henderson | G | 95 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.4 | 51.0 | 4 | FBref Prior G_service |
| Honest Ahanor | D | 90 | 0 | 0.02 | - | n/a | 0 | - | 7.2 | - | 0 | Prior D_service |
| Jaydee Canvot | D | 95 | 0 | 0.00 | 0.01 | 100% | 0 | 0.2 | 7.0 | 48.0 | 4 | FBref Prior D_service |
| Chris Richards | D | 95 | 0 | 0.00 | 0.02 | 100% | 1 | 0.5 | 7.2 | 49.8 | 4 | FBref Prior D_service |
| Adam Wharton | M | 95 | 0 | 0.06 | - | n/a | 1 | 0.0 | 7.7 | 52.5 | 4 | FBref Prior M_service |
| Quinten Timber | M | 78 | 0 | 0.06 | 0.01 | 541% | 1 | 0.5 | 7.1 | 32.5 | 2 | XGBoost M_service |
| Anan Khalaili | M | 70 | 0 | 0.10 | 0.18 | 47% | 1 | 0.8 | 7.3 | 55.5 | 4 | FBref Prior M_service |
| Tyrick Mitchell | M | 89 | 0 | 0.11 | 0.19 | 41% | 1 | 1.2 | 7.2 | 54.5 | 4 | FBref Prior M_service |
| Ismaïla Sarr | F | 90 | 0 | 0.27 | - | n/a | 3 | - | 7.4 | - | 0 | Prior F_service |
| Jørgen Strand Larsen | F | 49 | 0 | 0.17 | 0.10 | 76% | 2 | 0.7 | 7.2 | 68.0 | 3 | FBref Prior F_service |
| Jean-Philippe Mateta | F | 67 | 0 | 0.12 | 0.70 | 83% | 3 | 2.0 | 7.0 | 59.0 | 1 | FBref Prior F_service |
| Team totals | 0 | 1.02 | | 14 | | 7.3 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Walter Benítez | G | 4.4m | 11.0% | Market-value substitute prior |
| Daniel Muñoz | D | 14.4m | 23.5% | Market-value substitute prior |
| Chadi Riad | D | 14.3m | 23.2% | Market-value substitute prior |
| Dwight McNeil | M | 22.9m | 51.4% | Market-value substitute prior |
| Óscar Mingueza | M | 19.0m | 70.7% | Market-value substitute prior |
| Yéremy Pino | F | 31.6m | 54.7% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Ethan Ampadu vs Adam Wharton will be the pivotal duel. Total antagonist xG — Home: 0.90, Away: 0.69. Home danger zone: LW vs CB. Away danger zone: Shadow Striker vs CB + Mid Cover — Layered Defense.
Midfield — possession 49.7% / 50.3%, dominance ratio 0.48, key duel: Ethan Ampadu vs Adam Wharton
Harry Wilson xG 0.03Anton Stach xG 0.17Jean-Mattéo Bahoya xG 0.29Jean-Philippe Mateta xG 0.13Ismaïla Sarr xG 0.31Quinten Timber xG 0.07
Summary
Antagonist matchup analysis for Leeds vs Crystal Palace. Most likely scoreline: 1-0. Lambdas derived from 5 home and 5 away individual player duels.
Tactical parity between 4-3-3 and 3-4-2-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Ethan Ampadu vs Adam Wharton will be the pivotal duel. Total antagonist xG — Home: 0.90, Away: 0.69. Home danger zone: LW vs CB. Away danger zone: Shadow Striker vs CB + Mid Cover — Layered Defense.
Villarreal vs Levante
LaLiga · 20260920 · 5-4-1 vs 4-3-3
Predicted 1-2Winner: LevanteConfidence 0.57λ home/away 1.68 / 1.73Most-likely P 8.3%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Villarreal vs Levante. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 5-4-1 vs 4-3-3 (duel lambdas: Home 1.542, Away 1.097). Final calibrated Poisson lambdas: Home 1.677, Away 1.730 (most likely score 1-2).
Home Starters 10/11 XGBoost (10 FBref) Away Starters 8/11 XGBoost (0 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.54 / 1.10 Final λ (H/A) 1.68 / 1.73
FBref Team Priors: Villarreal (Gls/90: 1.3333333333333333, GA/90: 1.33 → E[λ]: 1.649) vs Levante UD (Gls/90: 1.6666666666666667, GA/90: 1.67 → E[λ]: 1.642) | 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 | Villarreal | Levante |
|---|
| League position | 17.00 | 11.00 |
| Points per game | 0.50 | 1.25 |
| Attack index | 1.25 | 1.04 |
| Defense (concede) index | 1.04 | 1.67 |
| Form multiplier | 0.96 | 1.04 |
| Standing λ multiplier | 1.10 | 1.26 |
Lambda decomposition (how the goal rates were built)
| Stage | Villarreal λ | Levante λ |
|---|
| FBref Squad Prior E[λ] (weight 0.228) | 1.649 | 1.642 |
| Pre-mismatch | 1.542 | 1.097 |
| Pre-competitive-floor | 1.677 | 1.072 |
| Pre-calibration (Micro-duels) | 1.542 | 1.097 |
| Final λ | 1.677 | 1.730 |
competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Truecompetitive underdog floor applied Truestage discount 1.0
Scoreline distribution & markets
| Top 5 scorelines | P |
|---|
| 1-2 | 8.3% |
| 1-1 | 9.7% |
| 2-1 | 8.1% |
| 2-2 | 7.0% |
| 0-1 | 5.7% |
| Home win | 37.5% |
| Draw | 22.8% |
| Away win | 39.7% |
| Over 2.5 | 66.2% |
| BTTS yes | 67.0% |
Pregame fair line — favorite Levante (39.7%), fair total 3.41, fair spread -0.05
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.
Villarreal — projected line vs season baseline (5-4-1)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Luiz Júnior | G | 91 | 0 | 0.00 | - | n/a | | 0.0 | 6.6 | 40.2 | 6 | XGBoost G_service |
| Renato Veiga | D | 95 | 0 | 0.00 | 0.02 | 100% | 1 | 0.3 | 7.3 | 45.3 | 6 | FBref Prior D_service |
| Carlos Romero | D | 95 | 0 | 0.00 | 0.01 | 100% | 1 | 0.4 | 7.1 | 40.4 | 5 | FBref Prior D_service |
| Santiago Mouriño | D | 83 | 0 | 0.00 | 0.03 | 100% | 1 | 0.3 | 6.9 | 33.5 | 6 | FBref Prior D_service |
| Pau Navarro | D | 90 | 0 | 0.04 | - | n/a | 0 | 0.0 | 7.0 | - | 6 | FBref Prior D_service |
| Logan Costa | D | 85 | 0 | 0.14 | 0.01 | 1417% | 1 | 0.2 | 6.9 | 11.0 | 6 | FBref Prior D_service |
| Alberto Moleiro | M | 56 | 0 | 0.18 | 0.04 | 393% | 2 | 0.3 | 7.6 | 45.8 | 6 | FBref Prior M_service |
| Pape Gueye | M | 90 | 0 | 0.16 | 0.02 | 874% | 1 | 0.7 | 7.5 | 47.8 | 6 | FBref Prior M_service |
| Tajon Buchanan | M | 39 | 0 | 0.24 | 0.09 | 156% | 2 | 0.7 | 7.0 | 44.3 | 6 | FBref Prior M_service |
| Ilias Akhomach | M | 14 | 0 | 0.15 | 0.00 | 6963% | 2 | 0.2 | 6.9 | 42.0 | 6 | FBref Prior M_service |
| Georges Mikautadze | F | 75 | 1 | 0.53 | 0.40 | 32% | 2 | 2.0 | 7.4 | 44.2 | 6 | FBref Prior F_service |
| Team totals | 1 | 1.68 | | 14 | | 7.1 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Péter Gulácsi | G | 4.4m | 10.9% | Market-value substitute prior |
| Alexander Freeman | D | 13.9m | 31.2% | Market-value substitute prior |
| Juan Foyth | D | 13.8m | 22.2% | Market-value substitute prior |
| Nathan-Dylan Saliba | M | 22.3m | 51.2% | Market-value substitute prior |
| Santi Comesaña | M | 22.2m | 31.2% | Market-value substitute prior |
| Tani Oluwaseyi | F | 15.9m | 72.3% | Market-value substitute prior |
Levante — 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 |
|---|
| Mathew Ryan | G | 88 | 0 | 0.00 | - | n/a | | 0.0 | 7.2 | 61.8 | 5 | XGBoost G_service |
| Axel Tape | D | 90 | 0 | 0.03 | - | n/a | 0 | 0.0 | 6.9 | - | 2 | Prior D_service |
| Manuel Sánchez | D | 89 | 0 | 0.00 | 0.00 | 100% | 0 | 0.2 | 6.6 | 54.8 | 5 | XGBoost D_service |
| Adrián de la Fuente | D | 89 | 0 | 0.19 | 0.09 | 127% | 1 | 1.2 | 6.7 | 59.2 | 5 | XGBoost D_service |
| Ifeanyi Ndukwe | D | 90 | 0 | 0.03 | - | n/a | 0 | 0.0 | 6.8 | - | 4 | Prior D_service |
| Enzo Bardeli | M | 69 | 0 | 0.08 | 0.01 | 579% | 1 | 0.4 | 6.8 | 54.0 | 5 | XGBoost M_service |
| Hugo Sotelo | M | 13 | 0 | 0.08 | 0.02 | 288% | 1 | 0.2 | 6.6 | 33.5 | 4 | XGBoost M_service |
| Thiago Fernández | M | 75 | 0 | 0.26 | 0.11 | 132% | 1 | 1.2 | 6.8 | 51.5 | 4 | XGBoost M_service |
| Etta Eyong | F | 22 | 0 | 0.14 | 0.02 | 501% | 2 | 0.5 | 6.9 | 34.2 | 4 | XGBoost F_service |
| Petar Ratkov | F | 90 | 1 | 0.45 | - | n/a | 2 | 0.0 | 7.0 | - | 2 | Prior F_service |
| Iván Romero | F | 71 | 1 | 0.36 | 0.33 | ok | 2 | 2.2 | 6.8 | 51.4 | 5 | XGBoost F_service |
| Team totals | 2 | 1.73 | | 10 | | 6.8 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Pablo Cunat Campos | G | 4.4m | 10.9% | Market-value substitute prior |
| Jeremy Toljan | D | 13.6m | 29.9% | Market-value substitute prior |
| Aïssa Mandi | D | 13.6m | 29.9% | Market-value substitute prior |
| Carlos Álvarez | M | 22.6m | 31.9% | Market-value substitute prior |
| Jon Ander Olasagasti | M | 22.0m | 49.9% | Market-value substitute prior |
| Kareem Tunde | F | 19.6m | 71.9% | Market-value substitute prior |
Antagonist / tactical summary
Home midfield dominates with ~53% possession expected (Alberto Moleiro vs Enzo Bardeli). Total antagonist xG — Home: 0.63, Away: 0.56. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Right Wing vs Left Wingback — Compact Flank.
Midfield — possession 52.9% / 47.1%, dominance ratio 0.60, key duel: Alberto Moleiro vs Enzo Bardeli
Georges Mikautadze xG 0.22Tajon Buchanan xG 0.12Ilias Akhomach xG 0.07Iván Romero xG 0.20Petar Ratkov xG 0.25Etta Eyong xG 0.08
Summary
Antagonist matchup analysis for Villarreal vs Levante. Most likely scoreline: 1-2. Lambdas derived from 6 home and 4 away individual player duels.
Tactical parity between 5-4-1 and 4-3-3. Match outcome driven primarily by individual player quality. Home midfield dominates with ~53% possession expected (Alberto Moleiro vs Enzo Bardeli). Total antagonist xG — Home: 0.63, Away: 0.56. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Right Wing vs Left Wingback — Compact Flank.
Bournemouth vs Liverpool
Premier League · 20260920 · 4-4-2 vs 3-4-2-1
Predicted 0-1Winner: LiverpoolConfidence 0.58λ home/away 1.14 / 1.61Most-likely P 10.2%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Bournemouth vs Liverpool. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-4-2 vs 3-4-2-1 (duel lambdas: Home 1.144, Away 1.609). Final calibrated Poisson lambdas: Home 1.144, Away 1.609 (most likely score 0-1).
Home Starters 9/11 XGBoost (10 FBref) Away Starters 9/11 XGBoost (0 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.14 / 1.61 Final λ (H/A) 1.14 / 1.61
FBref Team Priors: Bournemouth (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 1.481) vs Liverpool FC (Gls/90: 2.0, GA/90: 2.0 → E[λ]: 1.481) | 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 | Liverpool |
|---|
| League position | 15.00 | 6.00 |
| Points per game | 0.67 | 1.67 |
| Attack index | 1.67 | 1.25 |
| Defense (concede) index | 0.83 | 2.08 |
| Form multiplier | 0.95 | 1.05 |
| Standing λ multiplier | 1.13 | 1.50 |
Lambda decomposition (how the goal rates were built)
| Stage | Bournemouth λ | Liverpool λ |
|---|
| FBref Squad Prior E[λ] (weight 0.192) | 1.481 | 1.481 |
| Pre-competitive-floor | 1.144 | 1.609 |
| Pre-calibration (Micro-duels) | 1.144 | 1.609 |
| Final λ | 1.144 | 1.609 |
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 |
|---|
| 0-1 | 10.2% |
| 1-1 | 11.8% |
| 1-2 | 9.4% |
| 0-2 | 8.2% |
| 1-0 | 7.2% |
| Home win | 27.0% |
| Draw | 24.9% |
| Away win | 48.1% |
| Over 2.5 | 51.9% |
| BTTS yes | 54.6% |
Pregame fair line — favorite Liverpool (48.1%), fair total 2.75, fair spread -0.47
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 (4-4-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Đorđe Petrović | G | 95 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.5 | 53.0 | 4 | FBref Prior G_service |
| Adrien Truffert | D | 95 | 0 | 0.00 | 0.00 | 100% | 0 | 0.2 | 7.1 | 51.2 | 4 | FBref Prior D_service |
| António Silva | D | 95 | 0 | 0.00 | 0.07 | 100% | 1 | 0.5 | 7.2 | 52.5 | 4 | FBref Prior D_service |
| James Hill | D | 95 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.1 | 51.8 | 4 | FBref Prior D_service |
| Bafodé Diakité | D | 90 | 0 | 0.03 | - | n/a | 0 | 0.0 | 7.1 | - | 3 | FBref Prior D_service |
| Rayan | M | 95 | 0 | 0.17 | 0.12 | 38% | 1 | 1.8 | 7.5 | 46.8 | 4 | FBref Prior M_service |
| Alex Scott | M | 95 | 0 | 0.14 | 0.08 | 75% | 2 | 1.0 | 7.7 | 55.2 | 4 | FBref Prior M_service |
| Tyler Adams | M | 39 | 0 | 0.07 | - | n/a | 1 | 0.0 | 7.2 | 49.8 | 4 | FBref Prior M_service |
| Justin Kluivert | M | 81 | 0 | 0.12 | 0.19 | 38% | 1 | 1.8 | 7.1 | 49.2 | 4 | FBref Prior M_service |
| Eli Junior Kroupi | F | 90 | 0 | 0.34 | - | n/a | 3 | - | 7.5 | - | 0 | Prior F_service |
| Evanilson | F | 88 | 0 | 0.19 | 0.18 | ok | 2 | 1.5 | 7.2 | 50.5 | 4 | FBref Prior F_service |
| Team totals | 0 | 1.14 | | 12 | | 7.3 | | |
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 |
| Veljko Milosavljević | D | 14.3m | 33.1% | Market-value substitute prior |
| Julio Soler | D | 13.8m | 31.0% | Market-value substitute prior |
| Marcus Tavernier | M | 22.9m | 32.5% | Market-value substitute prior |
| Amine Adli | M | 22.7m | 53.0% | Market-value substitute prior |
| Álvaro Rodriguez | F | 17.0m | 72.5% | Market-value substitute prior |
Liverpool — 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 |
|---|
| Giorgi Mamardashvili | G | 90 | 0 | 0.00 | - | n/a | | 0.0 | 7.1 | - | 4 | Prior G_service |
| Jérémy Jacquet | D | 87 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.6 | 45.4 | 5 | XGBoost D_service |
| Miloš Kerkez | D | 90 | 0 | 0.00 | 0.00 | 100% | 1 | 0.4 | 7.1 | 38.6 | 5 | XGBoost D_service |
| Jeremie Frimpong | D | 87 | 0 | 0.00 | 0.00 | 100% | 1 | 0.2 | 7.2 | 30.3 | 4 | XGBoost D_service |
| Florian Wirtz | M | 90 | 0 | 0.20 | 0.13 | 51% | 1 | 1.4 | 7.8 | 38.6 | 5 | XGBoost M_service |
| Dominik Szoboszlai | M | 95 | 0 | 0.14 | 0.23 | 40% | | 1.8 | 7.7 | 43.4 | 5 | XGBoost M_service |
| Ryan Gravenberch | M | 46 | 0 | 0.06 | 0.02 | 257% | 1 | 0.5 | 7.6 | 50.2 | 4 | XGBoost M_service |
| Alexis Mac Allister | M | 70 | 0 | 0.05 | 0.02 | 197% | 2 | 0.4 | 7.7 | 41.8 | 5 | XGBoost M_service |
| Alexander Isak | F | 91 | 1 | 0.44 | 0.46 | ok | 3 | 1.8 | 8.3 | 45.8 | 5 | XGBoost F_service |
| Bradley Barcola | F | 29 | 0 | 0.15 | - | n/a | 2 | 0.0 | 7.6 | 33.0 | 2 | XGBoost F_service |
| Hugo Ekitiké | F | 90 | 0 | 0.44 | - | n/a | 4 | - | 7.5 | - | 1 | Prior F_service |
| Team totals | 1 | 1.61 | | 15 | | 7.6 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Alisson | G | 5.0m | 13.7% | Market-value substitute prior |
| Conor Bradley | D | 14.6m | 34.5% | Market-value substitute prior |
| Giovanni Leoni | D | 14.6m | 34.5% | Market-value substitute prior |
| Cody Gakpo | M | 23.6m | 57.3% | Market-value substitute prior |
| Trey Nyoni | M | 22.2m | 51.1% | Market-value substitute prior |
| Rio Ngumoha | F | 26.4m | 60.8% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Rayan vs Florian Wirtz will be the pivotal duel. Total antagonist xG — Home: 0.78, Away: 0.82. 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 48.5% / 51.5%, dominance ratio 0.46, key duel: Rayan vs Florian Wirtz
Rayan xG 0.13Alex Scott xG 0.11Tyler Adams xG 0.05Alexander Isak xG 0.39Bradley Barcola xG 0.13Florian Wirtz xG 0.19
Summary
Antagonist matchup analysis for Bournemouth vs Liverpool. Most likely scoreline: 0-1. Lambdas derived from 6 home and 5 away individual player duels.
Tactical parity between 4-4-2 and 3-4-2-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Rayan vs Florian Wirtz will be the pivotal duel. Total antagonist xG — Home: 0.78, Away: 0.82. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.
Getafe vs Málaga
LaLiga · 20260920 · 3-4-2-1 vs 3-5-2
Predicted 1-0Winner: GetafeConfidence 0.57λ home/away 1.26 / 1.02Most-likely P 12.8%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Getafe vs Málaga. 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 3-5-2 (duel lambdas: Home 1.260, Away 0.599). Final calibrated Poisson lambdas: Home 1.260, Away 1.020 (most likely score 1-0).
Home Starters 10/11 XGBoost (10 FBref) Away Starters 8/11 XGBoost (0 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.26 / 0.60 Final λ (H/A) 1.26 / 1.02
FBref Team Priors: Getafe (Gls/90: 0.3333333333333333, GA/90: 0.33 → E[λ]: 0.2) vs Málaga CF (Gls/90: 0.3333333333333333, GA/90: 0.33 → 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 | Getafe | Málaga |
|---|
| League position | 15.00 | 18.00 |
| Points per game | 1.00 | 0.50 |
| Attack index | 0.42 | 0.21 |
| Defense (concede) index | 1.46 | 1.04 |
| Form multiplier | 1.03 | 0.97 |
| Standing λ multiplier | 0.83 | 0.63 |
Lambda decomposition (how the goal rates were built)
| Stage | Getafe λ | Málaga λ |
|---|
| FBref Squad Prior E[λ] (weight 0.228) | 0.200 | 0.200 |
| Pre-competitive-floor | 1.260 | 0.650 |
| Pre-calibration (Micro-duels) | 1.260 | 0.599 |
| Final λ | 1.260 | 1.020 |
competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Falsecompetitive underdog floor applied Truestage discount 1.0
Scoreline distribution & markets
| Top 5 scorelines | P |
|---|
| 1-0 | 12.8% |
| 1-1 | 13.3% |
| 0-0 | 10.3% |
| 0-1 | 10.3% |
| 2-1 | 8.3% |
| Home win | 41.7% |
| Draw | 28.5% |
| Away win | 29.8% |
| Over 2.5 | 39.9% |
| BTTS yes | 45.9% |
Pregame fair line — favorite Getafe (41.7%), fair total 2.28, fair spread 0.24
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.
Getafe — 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 |
|---|
| David Soria | G | 88 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.0 | 56.2 | 4 | FBref Prior G_service |
| Andrés García | D | 77 | 0 | 0.00 | 0.03 | 100% | 1 | 0.8 | 6.5 | 53.0 | 4 | FBref Prior D_service |
| Abdel Abqar | D | 1 | 0 | 0.00 | - | n/a | 0 | 0.0 | 4.9 | - | 2 | FBref Prior D_service |
| Zaid Romero | D | 81 | 0 | 0.00 | 0.01 | 100% | 0 | 0.2 | 6.5 | 49.5 | 4 | FBref Prior D_service |
| Mario Martín | M | 91 | 0 | 0.09 | 0.03 | 169% | 2 | 1.0 | 6.8 | 33.0 | 2 | FBref Prior M_service |
| Orel Mangala | M | 89 | 0 | 0.04 | 0.03 | 36% | 2 | 0.5 | 6.9 | 49.5 | 4 | FBref Prior M_service |
| Ramón Terrats | M | 88 | 0 | 0.17 | 0.14 | ok | 3 | 2.5 | 6.8 | 52.2 | 4 | FBref Prior M_service |
| Francho Serrano | M | 81 | 0 | 0.05 | - | n/a | 1 | 0.0 | 6.7 | 32.8 | 4 | FBref Prior M_service |
| Christantus Uche | F | 90 | 0 | 0.22 | - | n/a | 3 | - | 7.0 | - | 1 | Prior F_service |
| Martín Satriano | F | 83 | 1 | 0.37 | 0.15 | 143% | 3 | 2.0 | 6.9 | 51.5 | 4 | FBref Prior F_service |
| Enes Ünal | F | 90 | 0 | 0.30 | 0.27 | ok | 3 | 1.0 | 6.7 | 50.5 | 4 | FBref Prior F_service |
| Team totals | 1 | 1.26 | | 19 | | 6.6 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Jiří Letáček | G | 4.4m | 10.9% | Market-value substitute prior |
| Sebastián Boselli | D | 10.5m | 60.2% | Market-value substitute prior |
| Djené | D | 13.6m | 21.7% | Market-value substitute prior |
| Nemanja Gudelj | M | 18.1m | 69.2% | Market-value substitute prior |
| Oscar Lopez | M | 22.0m | 49.9% | Market-value substitute prior |
| Borja Mayoral | F | 13.1m | 72.2% | Market-value substitute prior |
Málaga — projected line vs season baseline (3-5-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Alfonso Herrero | G | 88 | 0 | 0.00 | - | n/a | | 0.0 | 7.1 | 59.8 | 5 | XGBoost G_service |
| Adam Aznou | D | 90 | 0 | 0.02 | - | n/a | 0 | 0.0 | 6.9 | - | 2 | Prior D_service |
| Fernando Calero | D | 90 | 0 | 0.02 | - | n/a | 0 | - | 6.8 | - | 1 | Prior D_service |
| Jose Salinas | D | 45 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.5 | 39.2 | 5 | XGBoost D_service |
| Jens-Lys Cajuste | M | 9 | 0 | 0.05 | - | n/a | 2 | 0.0 | 6.4 | 33.5 | 2 | XGBoost M_service |
| David Larrubia | M | 75 | 0 | 0.19 | 0.09 | 123% | | 1.5 | 6.8 | 44.8 | 6 | XGBoost M_service |
| Pablo Martínez | M | 51 | 0 | 0.04 | 0.00 | 1081% | 2 | 0.2 | 6.8 | 33.0 | 4 | XGBoost M_service |
| Izan Merino | M | 72 | 0 | 0.04 | 0.01 | 333% | 1 | 0.2 | 6.7 | 53.6 | 5 | XGBoost M_service |
| Juan Cruz | M | 41 | 0 | 0.09 | 0.01 | 1016% | 1 | 0.4 | 6.7 | 51.6 | 5 | XGBoost M_service |
| Chupete | F | 91 | 0 | 0.27 | 0.58 | 54% | 2 | 2.4 | 6.8 | 51.2 | 5 | XGBoost F_service |
| Adrián Niño | F | 90 | 0 | 0.17 | - | n/a | 2 | - | 6.9 | - | 1 | Prior F_service |
| Team totals | 0 | 1.02 | | 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 |
| Diego Murillo | D | 13.6m | 29.9% | Market-value substitute prior |
| Juan Berrocal | D | 13.6m | 29.9% | Market-value substitute prior |
| Dani Lorenzo | M | 22.0m | 49.9% | Market-value substitute prior |
| Joaquín Muñoz | M | 22.0m | 49.9% | Market-value substitute prior |
| Julen Lobete | F | 25.3m | 55.9% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Mario Martín vs Jens-Lys Cajuste will be the pivotal duel. Total antagonist xG — Home: 1.25, Away: 0.57. 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 50.1% / 49.9%, dominance ratio 0.50, key duel: Mario Martín vs Jens-Lys Cajuste
Ramón Terrats xG 0.17Mario Martín xG 0.10Orel Mangala xG 0.04Izan Merino xG 0.04Jens-Lys Cajuste xG 0.05Pablo Martínez xG 0.03
Summary
Antagonist matchup analysis for Getafe vs Málaga. Most likely scoreline: 1-0. Lambdas derived from 7 home and 6 away individual player duels.
Tactical parity between 3-4-2-1 and 3-5-2. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Mario Martín vs Jens-Lys Cajuste will be the pivotal duel. Total antagonist xG — Home: 1.25, Away: 0.57. Home danger zone: Shadow Striker vs Centre Back — Half-Space Threat. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.
Fulham vs Man Utd
Premier League · 20260920 · 3-5-2 vs 3-4-2-1
Predicted 1-3Winner: Man UtdConfidence 0.66λ home/away 1.15 / 3.01Most-likely P 8.2%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Fulham vs Man Utd. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 3-5-2 vs 3-4-2-1 (duel lambdas: Home 1.179, Away 2.750). Final calibrated Poisson lambdas: Home 1.149, Away 3.006 (most likely score 1-3).
Home Starters 11/11 XGBoost (10 FBref) Away Starters 9/11 XGBoost (0 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.18 / 2.75 Final λ (H/A) 1.15 / 3.01
FBref Team Priors: Fulham (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 1.852) vs Manchester United (Gls/90: 2.0, GA/90: 2.5 → E[λ]: 1.481) | 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 | Fulham | Man Utd |
|---|
| League position | 19.00 | 11.00 |
| Points per game | 0.00 | 1.33 |
| Attack index | 1.67 | 1.46 |
| Defense (concede) index | 1.25 | 2.92 |
| Form multiplier | 0.95 | 1.05 |
| Standing λ multiplier | 1.34 | 1.65 |
Lambda decomposition (how the goal rates were built)
| Stage | Fulham λ | Man Utd λ |
|---|
| FBref Squad Prior E[λ] (weight 0.192) | 1.852 | 1.481 |
| Pre-mismatch | 1.179 | 2.750 |
| Pre-calibration (Micro-duels) | 1.179 | 2.750 |
| Final λ | 1.149 | 3.006 |
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 |
|---|
| 1-3 | 8.2% |
| 1-2 | 8.2% |
| 0-3 | 7.1% |
| 0-2 | 7.1% |
| 1-4 | 6.1% |
| Home win | 11.3% |
| Draw | 14.0% |
| Away win | 74.7% |
| Over 2.5 | 78.3% |
| BTTS yes | 64.9% |
Pregame fair line — favorite Man Utd (74.7%), fair total 4.16, fair spread -1.86
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.
Fulham — projected line vs season baseline (3-5-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Bernd Leno | G | 89 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.7 | 46.8 | 4 | FBref Prior G_service |
| Calvin Bassey | D | 95 | 0 | 0.00 | 0.00 | 100% | 0 | 0.2 | 7.4 | 49.8 | 4 | FBref Prior D_service |
| Antonee Robinson | D | 88 | 0 | 0.00 | 0.05 | 100% | 1 | 0.8 | 7.0 | 47.8 | 4 | FBref Prior D_service |
| Joachim Andersen | D | 94 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.2 | 46.3 | 3 | FBref Prior D_service |
| Hugo Larsson | M | 12 | 0 | 0.10 | 0.13 | ok | 2 | 1.0 | 7.3 | 73.0 | 1 | XGBoost M_service |
| Oscar Bobb | M | 75 | 0 | 0.22 | 0.08 | 171% | 2 | 0.8 | 7.4 | 51.8 | 4 | FBref Prior M_service |
| Kevin | M | 15 | 0 | 0.06 | - | n/a | 1 | 0.0 | 7.2 | 43.0 | 3 | FBref Prior M_service |
| Sander Berge | M | 65 | 0 | 0.05 | 0.01 | 381% | 1 | 0.3 | 7.3 | 66.3 | 3 | FBref Prior M_service |
| Joshua King | M | 95 | 0 | 0.22 | 0.24 | ok | 2 | 1.8 | 7.4 | 52.5 | 4 | FBref Prior M_service |
| Gonzalo García | F | 95 | 1 | 0.21 | 0.36 | 41% | 3 | 2.8 | 7.3 | 49.8 | 4 | FBref Prior F_service |
| Rodrigo Muniz | F | 10 | 0 | 0.13 | 0.13 | ok | 2 | 1.0 | 6.3 | 42.7 | 3 | FBref Prior F_service |
| Team totals | 1 | 1.15 | | 16 | | 7.1 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Benjamin Lecomte | G | 4.4m | 10.9% | Market-value substitute prior |
| Ryan Sessegnon | D | 18.0m | 61.8% | Market-value substitute prior |
| David Affengruber | D | 14.3m | 33.0% | Market-value substitute prior |
| Alex Iwobi | M | 22.8m | 32.3% | Market-value substitute prior |
| Emile Smith Rowe | M | 18.4m | 70.6% | Market-value substitute prior |
| Jonah Kusi-Asare | F | 25.3m | 55.9% | Market-value substitute prior |
Man Utd — 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 |
|---|
| Senne Lammens | G | 95 | 0 | 0.00 | - | n/a | | 0.0 | 7.4 | 68.0 | 3 | XGBoost G_service |
| Lisandro Martínez | D | 95 | 0 | 0.00 | 0.01 | 100% | 1 | 0.3 | 7.4 | 44.3 | 3 | XGBoost D_service |
| Leny Yoro | D | 16 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.3 | 42.3 | 3 | XGBoost D_service |
| Ayden Heaven | D | 95 | 0 | 0.24 | 0.04 | 486% | 1 | 0.7 | 7.2 | 21.0 | 3 | XGBoost D_service |
| Kobbie Mainoo | M | 69 | 0 | 0.11 | 0.07 | 62% | 1 | 1.3 | 7.9 | 73.7 | 3 | XGBoost M_service |
| Carlos Baleba | M | 90 | 0 | 0.25 | - | n/a | 2 | - | 7.5 | - | 0 | Prior M_service |
| Amad Diallo | M | 90 | 0 | 0.24 | - | n/a | 2 | - | 7.4 | - | 0 | Prior M_service |
| Youri Tielemans | M | 90 | 0 | 0.15 | 0.18 | ok | 1 | 2.0 | 7.9 | 71.7 | 3 | XGBoost M_service |
| Matheus Cunha | F | 85 | 1 | 0.61 | 0.16 | 291% | 3 | 2.3 | 7.5 | 63.7 | 3 | XGBoost F_service |
| Benjamin Šeško | F | 20 | 1 | 0.37 | 0.11 | 248% | 3 | 1.0 | 7.7 | 66.0 | 3 | XGBoost F_service |
| Bryan Mbeumo | F | 95 | 1 | 0.78 | 0.58 | 35% | 3 | 2.8 | 7.7 | 53.0 | 4 | XGBoost F_service |
| Team totals | 3 | 3.01 | | 19 | | 7.5 | | |
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 |
| Matthijs de Ligt | D | 14.7m | 35.2% | Market-value substitute prior |
| Diogo Dalot | D | 14.6m | 34.6% | Market-value substitute prior |
| Andrey Santos | M | 23.3m | 56.0% | Market-value substitute prior |
| Patrick Dorgu | M | 23.2m | 55.3% | Market-value substitute prior |
| Marcus Rashford | F | 26.6m | 61.9% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Hugo Larsson vs Kobbie Mainoo will be the pivotal duel. Total antagonist xG — Home: 0.61, Away: 1.42. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Shadow Striker vs Centre Back — Half-Space Threat.
Midfield — possession 48.9% / 51.1%, dominance ratio 0.47, key duel: Hugo Larsson vs Kobbie Mainoo
Kevin xG 0.05Oscar Bobb xG 0.16Hugo Larsson xG 0.08Kobbie Mainoo xG 0.06Carlos Baleba xG 0.16Amad Diallo xG 0.15
Summary
Antagonist matchup analysis for Fulham vs Man Utd. Most likely scoreline: 1-3. Lambdas derived from 6 home and 7 away individual player duels.
Tactical parity between 3-5-2 and 3-4-2-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Hugo Larsson vs Kobbie Mainoo will be the pivotal duel. Total antagonist xG — Home: 0.61, Away: 1.42. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Shadow Striker vs Centre Back — Half-Space Threat.
Atl. Madrid vs Real Madrid
LaLiga · 20260920 · 3-5-2 vs 3-4-2-1
Predicted 1-2Winner: Real MadridConfidence 0.62λ home/away 1.99 / 2.67Most-likely P 6.7%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Atl. Madrid vs Real Madrid. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 3-5-2 vs 3-4-2-1 (duel lambdas: Home 1.531, Away 2.667). Final calibrated Poisson lambdas: Home 1.993, Away 2.667 (most likely score 1-2).
Home Starters 11/11 XGBoost (10 FBref) Away Starters 10/11 XGBoost (10 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 1.53 / 2.67 Final λ (H/A) 1.99 / 2.67
FBref Team Priors: Atlético Madrid (Gls/90: 2.333333333333333, GA/90: 2.33 → E[λ]: 3.5) vs Real Madrid (Gls/90: 3.0, GA/90: 3.33 → 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 | Atl. Madrid | Real Madrid |
|---|
| League position | 6.00 | 3.00 |
| Points per game | 1.75 | 2.25 |
| Attack index | 1.46 | 2.08 |
| Defense (concede) index | 0.62 | 1.25 |
| Form multiplier | 0.97 | 1.03 |
| Standing λ multiplier | 0.96 | 1.50 |
Lambda decomposition (how the goal rates were built)
| Stage | Atl. Madrid λ | Real Madrid λ |
|---|
| FBref Squad Prior E[λ] (weight 0.228) | 3.500 | 3.500 |
| Pre-competitive-floor | 1.531 | 2.667 |
| Pre-calibration (Micro-duels) | 1.531 | 2.667 |
| Final λ | 1.993 | 2.667 |
competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Falsecompetitive underdog floor applied Truestage discount 1.0
Scoreline distribution & markets
| Top 5 scorelines | P |
|---|
| 1-2 | 6.7% |
| 2-2 | 6.7% |
| 1-3 | 6.0% |
| 2-3 | 5.9% |
| 1-1 | 5.1% |
| Home win | 28.9% |
| Draw | 18.3% |
| Away win | 52.8% |
| Over 2.5 | 84.4% |
| BTTS yes | 80.4% |
Pregame fair line — favorite Real Madrid (52.8%), fair total 4.66, fair spread -0.67
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.
Atl. Madrid — projected line vs season baseline (3-5-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Jan Oblak | G | 92 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.1 | 54.4 | 5 | FBref Prior G_service |
| Cristian Romero | D | 34 | 0 | 0.34 | 0.06 | 488% | 1 | 0.3 | 7.3 | 22.3 | 3 | FBref Prior D_service |
| Dávid Hancko | D | 95 | 0 | 0.00 | 0.01 | 100% | 0 | 0.6 | 7.4 | 58.8 | 5 | FBref Prior D_service |
| Marc Pubill | D | 88 | 0 | 0.00 | 0.03 | 100% | 0 | 0.4 | 7.2 | 42.0 | 5 | FBref Prior D_service |
| Pablo Barrios | M | 81 | 0 | 0.12 | 0.03 | 297% | 2 | 0.6 | 7.8 | 58.6 | 5 | FBref Prior M_service |
| Morten Hjulmand | M | 67 | 0 | 0.09 | 0.00 | 2260% | 1 | 0.2 | 7.6 | 55.4 | 5 | FBref Prior M_service |
| Giuliano Simeone | M | 74 | 0 | 0.17 | 0.06 | 167% | 1 | 1.2 | 7.4 | 52.6 | 5 | FBref Prior M_service |
| Alex Baena | M | 65 | 0 | 0.21 | 0.10 | 122% | 2 | 1.4 | 7.6 | 59.2 | 5 | FBref Prior M_service |
| Kang-in Lee | M | 61 | 0 | 0.17 | 0.08 | 101% | 1 | 2.2 | 7.4 | 57.0 | 5 | XGBoost M_service |
| Julián Alvarez | F | 31 | 0 | 0.17 | 0.09 | 84% | 3 | 0.7 | 7.7 | 43.7 | 3 | FBref Prior F_service |
| Ademola Lookman | F | 80 | 1 | 0.34 | 0.09 | 304% | 3 | 1.4 | 7.4 | 54.8 | 5 | FBref Prior F_service |
| Team totals | 1 | 1.99 | | 16 | | 7.4 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Juan Musso | G | 4.4m | 10.9% | Market-value substitute prior |
| Marcos Llorente | D | 25.7m | 49.0% | Market-value substitute prior |
| Robin Le Normand | D | 14.3m | 23.2% | Market-value substitute prior |
| Rodrigo Mendoza | M | 13.6m | 45.2% | Market-value substitute prior |
| Alejandro Grimaldo | M | 22.7m | 53.3% | Market-value substitute prior |
| Jonathan David | F | 26.4m | 60.8% | Market-value substitute prior |
Real Madrid — 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 |
|---|
| Thibaut Courtois | G | 92 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.1 | 38.0 | 7 | FBref Prior G_service |
| Dean Huijsen | D | 95 | 0 | 0.00 | 0.08 | 100% | 1 | 0.7 | 7.7 | 41.6 | 7 | FBref Prior D_service |
| Trent Alexander-Arnold | D | 46 | 0 | 0.00 | 0.01 | 100% | 1 | 0.3 | 7.3 | 30.6 | 7 | FBref Prior D_service |
| Marc Cucurella | D | 69 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.4 | 46.5 | 6 | FBref Prior D_service |
| Jude Bellingham | M | 94 | 0 | 0.23 | 0.12 | 89% | 3 | 1.6 | 7.9 | 46.7 | 7 | FBref Prior M_service |
| Federico Valverde | M | 95 | 0 | 0.09 | 0.04 | 153% | 1 | 0.7 | 7.7 | 48.0 | 6 | FBref Prior M_service |
| Arda Güler | M | 59 | 0 | 0.13 | 0.16 | ok | 2 | 0.9 | 7.7 | 43.4 | 7 | FBref Prior M_service |
| Aurélien Tchouaméni | M | 90 | 0 | 0.23 | - | n/a | 2 | - | 7.5 | - | 2 | Prior M_service |
| Kylian Mbappé | F | 95 | 2 | 0.90 | 0.86 | ok | 5 | 4.4 | 8.5 | 45.0 | 7 | FBref Prior F_service |
| Vinícius Júnior | F | 95 | 0 | 0.52 | 0.27 | 92% | 3 | 1.4 | 7.9 | 36.5 | 8 | FBref Prior F_service |
| Yan Diomande | F | 24 | 0 | 0.21 | - | n/a | 2 | 0.0 | 7.6 | 44.3 | 6 | FBref Prior F_service |
| Team totals | 2 | 2.67 | | 22 | | 7.7 | | |
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 |
| Álvaro Carreras | D | 15.1m | 24.9% | Market-value substitute prior |
| Ibrahima Konaté | D | 15.0m | 24.7% | Market-value substitute prior |
| Eduardo Camavinga | M | 17.2m | 59.4% | Market-value substitute prior |
| Brahim Díaz | M | 17.6m | 52.6% | Market-value substitute prior |
| Rodrygo | F | 26.8m | 62.7% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Pablo Barrios vs Jude Bellingham will be the pivotal duel. Total antagonist xG — Home: 0.63, Away: 1.44. 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 49.3% / 50.7%, dominance ratio 0.48, key duel: Pablo Barrios vs Jude Bellingham
Giuliano Simeone xG 0.09Alex Baena xG 0.10Pablo Barrios xG 0.06Federico Valverde xG 0.06Aurélien Tchouaméni xG 0.16Jude Bellingham xG 0.15
Summary
Antagonist matchup analysis for Atl. Madrid vs Real Madrid. Most likely scoreline: 1-2. Lambdas derived from 6 home and 7 away individual player duels.
Tactical parity between 3-5-2 and 3-4-2-1. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Pablo Barrios vs Jude Bellingham will be the pivotal duel. Total antagonist xG — Home: 0.63, Away: 1.44. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.
Valencia vs Real Sociedad
LaLiga · 20260920 · 4-2-3-1 vs 5-3-2
Predicted 1-0Winner: ValenciaConfidence 0.61λ home/away 1.53 / 0.90Most-likely P 13.4%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Valencia vs Real Sociedad. Player statistics regressed using position-specific XGBoost models (G, D, M, F) with 2026-2027 FBref priors. Micro-duels resolved across 4-2-3-1 vs 5-3-2 (duel lambdas: Home 0.591, Away 0.898). Final calibrated Poisson lambdas: Home 1.533, Away 0.898 (most likely score 1-0).
Home Starters 9/11 XGBoost (7 FBref) Away Starters 11/11 XGBoost (10 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 0.59 / 0.90 Final λ (H/A) 1.53 / 0.90
FBref Team Priors: Valencia (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 | Valencia | Real Sociedad |
|---|
| League position | 20.00 | 9.00 |
| Points per game | 0.25 | 1.40 |
| Attack index | 0.21 | 0.83 |
| Defense (concede) index | 1.11 | 1.87 |
| Form multiplier | 0.95 | 1.05 |
| Standing λ multiplier | 0.63 | 1.21 |
Lambda decomposition (how the goal rates were built)
| Stage | Valencia λ | Real Sociedad λ |
|---|
| FBref Squad Prior E[λ] (weight 0.228) | 0.200 | 0.200 |
| Pre-competitive-floor | 0.650 | 0.898 |
| Pre-calibration (Micro-duels) | 0.591 | 0.898 |
| Final λ | 1.533 | 0.898 |
competition calibration scale 1.0competition observed total goals 0.0quality gap applied Truemismatch score pressure applied Falsecompetitive underdog floor applied Truestage discount 1.0
Scoreline distribution & markets
| Top 5 scorelines | P |
|---|
| 1-0 | 13.4% |
| 1-1 | 12.2% |
| 2-0 | 10.3% |
| 2-1 | 9.3% |
| 0-0 | 8.9% |
| Home win | 52.1% |
| Draw | 25.9% |
| Away win | 22.0% |
| Over 2.5 | 43.8% |
| BTTS yes | 46.6% |
Pregame fair line — favorite Valencia (52.1%), fair total 2.43, fair spread 0.64
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.
Valencia — 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 |
|---|
| Stole Dimitrievski | G | 88 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.8 | 43.2 | 6 | FBref Prior G_service |
| Arnau Martínez | D | 87 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.6 | 30.5 | 4 | FBref Prior D_service |
| César Tárrega | D | 77 | 0 | 0.00 | 0.02 | 100% | 0 | 0.5 | 6.8 | 44.8 | 6 | FBref Prior D_service |
| Justin De Haas | D | 45 | 0 | 0.00 | - | n/a | 0 | 0.0 | 6.7 | 33.0 | 2 | FBref Prior D_service |
| José Copete | D | 90 | 0 | 0.04 | - | n/a | 0 | - | 6.8 | - | 1 | Prior D_service |
| Javier Guerra | M | 80 | 0 | 0.30 | 0.04 | 620% | 2 | 0.5 | 7.2 | 43.5 | 6 | FBref Prior M_service |
| Harvey Elliott | M | 38 | 0 | 0.16 | - | n/a | 1 | 0.0 | 7.0 | 22.0 | 3 | XGBoost M_service |
| Diego López | M | 90 | 0 | 0.26 | - | n/a | 1 | - | 7.0 | - | 1 | Prior M_service |
| Pepelu | M | 74 | 0 | 0.12 | 0.00 | 2371% | 1 | 0.1 | 6.9 | 38.3 | 7 | XGBoost M_service |
| Filip Ugrinić | M | 70 | 0 | 0.27 | 0.07 | 286% | 1 | 1.0 | 6.8 | 44.0 | 6 | FBref Prior M_service |
| Hugo Duro | F | 28 | 1 | 0.23 | - | n/a | 2 | 0.0 | 6.9 | 51.6 | 5 | FBref Prior F_service |
| Team totals | 1 | 1.53 | | 8 | | 6.9 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Kayne van Oevelen | G | 4.4m | 10.9% | Market-value substitute prior |
| José Luis Gayà | D | 13.6m | 21.8% | Market-value substitute prior |
| Jesús Vázquez | D | 13.6m | 21.8% | Market-value substitute prior |
| André Almeida | M | 22.1m | 50.6% | Market-value substitute prior |
| Guido Rodríguez | M | 22.0m | 30.8% | Market-value substitute prior |
| Umar Sadiq | F | 30.6m | 59.2% | Market-value substitute prior |
Real Sociedad — 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 |
|---|
| Álex Remiro | G | 90 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.1 | 52.2 | 6 | FBref Prior G_service |
| Mamadou Sarr | D | 61 | 0 | 0.00 | - | n/a | 1 | 0.0 | 6.9 | 17.8 | 4 | FBref Prior D_service |
| Jon Martin | D | 77 | 0 | 0.00 | 0.02 | 100% | 0 | 0.3 | 7.0 | 52.7 | 6 | FBref Prior D_service |
| Sergio Gómez | D | 88 | 0 | 0.00 | 0.02 | 100% | 0 | 0.5 | 6.7 | 56.8 | 6 | FBref Prior D_service |
| Jon Aramburu | D | 85 | 0 | 0.00 | 0.01 | 100% | 1 | 0.7 | 6.7 | 55.0 | 6 | FBref Prior D_service |
| Hector Fort | D | 9 | 0 | 0.00 | - | n/a | 1 | 0.0 | 5.6 | 22.7 | 3 | FBref Prior D_service |
| Jon Gorrotxategi | M | 18 | 0 | 0.08 | - | n/a | 2 | 0.0 | 6.7 | 33.2 | 4 | FBref Prior M_service |
| Takefusa Kubo | M | 53 | 0 | 0.09 | 0.09 | ok | 1 | 0.3 | 7.2 | 44.5 | 6 | FBref Prior M_service |
| Ander Barrenetxea | M | 56 | 0 | 0.09 | 0.04 | 127% | 1 | 0.7 | 7.0 | 45.0 | 6 | FBref Prior M_service |
| Mikel Oyarzabal | F | 72 | 0 | 0.23 | 0.08 | 188% | 2 | 1.2 | 7.1 | 55.7 | 6 | FBref Prior F_service |
| Orri Óskarsson | F | 71 | 0 | 0.29 | 0.36 | ok | 2 | 1.5 | 7.3 | 45.2 | 6 | XGBoost F_service |
| Team totals | 0 | 0.90 | | 12 | | 6.9 | | |
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 |
| Igor Zubeldia | D | 24.9m | 31.6% | Market-value substitute prior |
| Jon Pacheco | D | 13.7m | 30.4% | Market-value substitute prior |
| Luka Sučić | M | 26.6m | 41.1% | Market-value substitute prior |
| Beñat Turrientes | M | 20.5m | 41.1% | Market-value substitute prior |
| Gorka Carrera Zarranz | F | 25.3m | 56.1% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Javier Guerra vs Jon Gorrotxategi will be the pivotal duel. Total antagonist xG — Home: 0.39, Away: 0.51. Home danger zone: CAM vs CM. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.
Midfield — possession 49.4% / 50.6%, dominance ratio 0.49, key duel: Javier Guerra vs Jon Gorrotxategi
Filip Ugrinić xG 0.07Diego López xG 0.06Javier Guerra xG 0.11Mikel Oyarzabal xG 0.20Sergio Gómez xG 0.08Jon Aramburu xG 0.08
Summary
Antagonist matchup analysis for Valencia vs Real Sociedad. Most likely scoreline: 1-0. Lambdas derived from 5 home and 5 away individual player duels.
Tactical parity between 4-2-3-1 and 5-3-2. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Javier Guerra vs Jon Gorrotxategi will be the pivotal duel. Total antagonist xG — Home: 0.39, Away: 0.51. Home danger zone: CAM vs CM. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.
Man City vs Sunderland
Premier League · 20260920 · 4-2-3-1 vs 3-5-2
Predicted 2-0Winner: Man CityConfidence 0.66λ home/away 2.60 / 0.98Most-likely P 9.4%
Actual vs predicted: result not available yet (upcoming fixture).
XGBoost Reasoning & FBref Dataset TracingAntagonist + XGBoost match analysis for Man City 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 3-5-2 (duel lambdas: Home 2.330, Away 1.019). Final calibrated Poisson lambdas: Home 2.604, Away 0.983 (most likely score 2-0).
Home Starters 11/11 XGBoost (10 FBref) Away Starters 11/11 XGBoost (9 FBref) Bench Modeled H: 7 / A: 7 Duel λ (H/A) 2.33 / 1.02 Final λ (H/A) 2.60 / 0.98
FBref Team Priors: Manchester City (Gls/90: 3.0, GA/90: 3.0 → E[λ]: 2.222) vs Sunderland (Gls/90: 1.0, GA/90: 1.0 → E[λ]: 2.222) | 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 | Man City | Sunderland |
|---|
| League position | 1.00 | 12.00 |
| Points per game | 3.00 | 1.33 |
| Attack index | 2.92 | 0.62 |
| Defense (concede) index | 0.62 | 0.83 |
| Form multiplier | 1.06 | 0.94 |
| Standing λ multiplier | 1.29 | 0.77 |
Lambda decomposition (how the goal rates were built)
| Stage | Man City λ | Sunderland λ |
|---|
| FBref Squad Prior E[λ] (weight 0.192) | 2.222 | 2.222 |
| Pre-mismatch | 2.330 | 1.019 |
| Pre-calibration (Micro-duels) | 2.330 | 1.019 |
| Final λ | 2.604 | 0.983 |
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 |
|---|
| 2-0 | 9.4% |
| 2-1 | 9.2% |
| 3-0 | 8.2% |
| 3-1 | 8.0% |
| 1-0 | 7.2% |
| Home win | 72.1% |
| Draw | 16.0% |
| Away win | 11.9% |
| Over 2.5 | 69.5% |
| BTTS yes | 58.0% |
Pregame fair line — favorite Man City (72.1%), fair total 3.59, fair spread 1.62
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.
Man City — 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 |
|---|
| Gianluigi Donnarumma | G | 95 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.9 | 57.8 | 4 | FBref Prior G_service |
| Nico O'Reilly | D | 85 | 0 | 0.10 | 0.03 | 235% | 1 | 0.7 | 7.4 | 44.0 | 3 | FBref Prior D_service |
| Marc Guéhi | D | 95 | 0 | 0.00 | 0.22 | 100% | 1 | 0.5 | 7.7 | 58.5 | 4 | FBref Prior D_service |
| Joško Gvardiol | D | 94 | 0 | 0.00 | 0.05 | 100% | 1 | 0.5 | 7.4 | 53.8 | 4 | FBref Prior D_service |
| Rúben Dias | D | 95 | 0 | 0.00 | 0.02 | 100% | 0 | 0.2 | 7.7 | 57.2 | 4 | FBref Prior D_service |
| Elliot Anderson | M | 88 | 0 | 0.22 | 0.02 | 899% | 1 | 0.8 | 8.5 | 57.8 | 4 | FBref Prior M_service |
| Rayan Cherki | M | 65 | 0 | 0.31 | 0.14 | 127% | 2 | 1.5 | 7.9 | 58.0 | 4 | FBref Prior M_service |
| Enzo Fernández | M | 56 | 0 | 0.19 | 0.04 | 343% | 2 | 0.7 | 7.9 | 47.3 | 3 | XGBoost M_service |
| Ayyoub Bouaddi | M | 12 | 0 | 0.14 | - | n/a | 0 | 0.0 | 7.6 | 45.7 | 3 | FBref Prior M_service |
| Antoine Semenyo | M | 95 | 0 | 0.36 | 0.04 | 729% | 2 | 1.2 | 7.7 | 50.2 | 4 | FBref Prior M_service |
| Erling Haaland | F | 95 | 2 | 1.15 | 0.62 | 86% | 3 | 3.5 | 8.2 | 59.2 | 4 | FBref Prior F_service |
| Team totals | 2 | 2.60 | | 13 | | 7.8 | | |
Bench Candidates (7 modeled substitutes with empirical playing time)
| Player | Pos | Exp Min | App Prob | Substitute Modeling Details |
|---|
| Gerónimo Rulli | G | 4.5m | 11.5% | Market-value substitute prior |
| Matheus Nunes | D | 25.5m | 63.3% | Market-value substitute prior |
| Abdukodir Khusanov | D | 15.1m | 24.7% | Market-value substitute prior |
| Jérémy Doku | M | 23.7m | 57.7% | Market-value substitute prior |
| Phil Foden | M | 23.7m | 34.2% | Market-value substitute prior |
| Omar Marmoush | F | 26.9m | 63.1% | Market-value substitute prior |
Sunderland — projected line vs season baseline (3-5-2)
| Player | Pos | Min | Proj G | Proj xG | Season xG/g | | Proj Sh | Season Sh/g | Proj Rtg | Season Rtg | Rec | Model / FBref Prior |
|---|
| Robin Roefs | G | 95 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.7 | 51.2 | 4 | FBref Prior G_service |
| Daniel Ballard | D | 94 | 0 | 0.00 | 0.02 | 100% | 1 | 0.5 | 7.1 | 51.8 | 4 | FBref Prior D_service |
| Trai Hume | D | 68 | 0 | 0.00 | 0.01 | 100% | 1 | 0.7 | 6.9 | 68.7 | 3 | FBref Prior D_service |
| Kevin Danso | D | 93 | 0 | 0.00 | - | n/a | 0 | 0.0 | 7.0 | 24.3 | 3 | XGBoost D_service |
| Noah Sadiki | M | 91 | 0 | 0.14 | 0.09 | 52% | 2 | 1.2 | 7.5 | 51.8 | 4 | FBref Prior M_service |
| Habib Diarra | M | 15 | 0 | 0.09 | 0.01 | 568% | 3 | 0.5 | 7.4 | 67.5 | 2 | FBref Prior M_service |
| Malick Fofana | M | 17 | 0 | 0.09 | - | n/a | 1 | 0.0 | 7.2 | 67.0 | 1 | XGBoost M_service |
| Enzo Le Fée | M | 91 | 0 | 0.13 | 0.30 | 55% | 1 | 1.8 | 7.3 | 51.8 | 4 | FBref Prior M_service |
| Chris Rigg | M | 12 | 0 | 0.06 | - | n/a | 1 | 0.0 | 7.0 | 43.0 | 3 | FBref Prior M_service |
| Brian Brobbey | F | 80 | 0 | 0.19 | 0.15 | ok | 2 | 1.5 | 7.4 | 48.0 | 4 | FBref Prior F_service |
| Wilson Isidor | F | 26 | 0 | 0.18 | 0.50 | 63% | 5 | 1.7 | 7.1 | 68.0 | 3 | FBref Prior F_service |
| Team totals | 0 | 0.98 | | 20 | | 7.2 | | |
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 |
| Nordi Mukiele | D | 14.1m | 22.9% | Market-value substitute prior |
| Omar Alderete | D | 19.5m | 61.2% | Market-value substitute prior |
| Chemsdine Talbi | M | 27.9m | 71.0% | Market-value substitute prior |
| Dayann Methalie | M | 22.7m | 53.3% | Market-value substitute prior |
| Romaine Mundle | F | 12.0m | 72.3% | Market-value substitute prior |
Antagonist / tactical summary
Midfield is evenly contested — Elliot Anderson vs Noah Sadiki will be the pivotal duel. Total antagonist xG — Home: 1.09, Away: 0.49. 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 51.5% / 48.5%, dominance ratio 0.55, key duel: Elliot Anderson vs Noah Sadiki
Rayan Cherki xG 0.17Antoine Semenyo xG 0.20Enzo Fernández xG 0.08Brian Brobbey xG 0.17Malick Fofana xG 0.07Chris Rigg xG 0.04
Summary
Antagonist matchup analysis for Man City vs Sunderland. Most likely scoreline: 2-0. Lambdas derived from 5 home and 5 away individual player duels.
Tactical parity between 4-2-3-1 and 3-5-2. Match outcome driven primarily by individual player quality. Midfield is evenly contested — Elliot Anderson vs Noah Sadiki will be the pivotal duel. Total antagonist xG — Home: 1.09, Away: 0.49. Home danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle. Away danger zone: Striker vs Two Centre Backs — 1v2 Aerial/Physical Battle.
Generated 2026-09-20 13:31:36 UTC · Soccer prediction reporting