2025-2026 GLEAGUE Model Performance Analysis
Scope
All scored games in the selected league and season. AP Poll is excluded here.
Season
Comparing prediction accuracy across 543 games using multiple rating models.
7-day holdout coverage: 16/17 models .
Rolling Holdout Curves
Each point is a strict weekly holdout: train on all games before that week, test on that week. This first version uses a 21-day warmup, then 7-day holdouts stepped forward weekly.
Weekly strict holdout log loss. Lower is better. Showing 16 models across 22 windows. Click legend items to hide/show series.
Recent Window Winners
| Holdout | Best | Log Loss | Runner-up | Models |
|---|---|---|---|---|
| Apr 8 - Apr 10 | Log Adjusted | 0.567 | Adjusted Efficiency (0.568) | 16 |
| Apr 1 - Apr 7 | Log Adjusted | 0.563 | Adjusted Efficiency (0.563) | 16 |
| Mar 25 - Mar 31 | Home Team Baseline | 0.650 | Points Off/Def Recency (0.669) | 16 |
| Mar 18 - Mar 24 | Dynamic Bradley-Terry | 0.674 | Bradley-Terry (0.676) | 16 |
| Mar 11 - Mar 17 | Home Team Baseline | 0.601 | Points Off/Def (0.715) | 16 |
| Mar 4 - Mar 10 | Pythagorean | 0.624 | Avg Margin Baseline (0.633) | 16 |
| Feb 25 - Mar 3 | Dynamic Bradley-Terry | 0.591 | Elo (0.601) | 16 |
| Feb 18 - Feb 24 | Points Off/Def Recency | 0.692 | Margin Recency (0.695) | 16 |
Model Performance Leaderboard
Models ranked by strict holdout AUC when available (fallback: full-season AUC). Hover over column headers for explanations.
| # | Model | 7d Split | AUC | Acc | Brier | LogLoss | n | AUC 7d | Acc 7d | Brier 7d | n 7d |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Pythagorean Pythagorean Pythagorean win expectation from raw points scored and allowed. More → |
STRICT
4g
|
0.532 | 53.4% | 0.265 | 0.731 | 756 | 0.667 | 75.0% | 0.243 | 4 |
| 2 | Adjusted Efficiency Adjusted Efficiency Opponent-adjusted efficiency model with separate offensive and defensive components. More → |
STRICT
4g
|
0.541 | 53.0% | 0.301 | 0.861 | 756 | 0.667 | 75.0% | 0.227 | 4 |
| 3 | Log Adjusted Log Adjusted Log-scale adjusted efficiency model that downweights blowout leverage. More → |
STRICT
4g
|
0.541 | 53.0% | 0.301 | 0.861 | 756 | 0.667 | 75.0% | 0.227 | 4 |
| 4 | Avg Margin Baseline Avg Margin Baseline Predict from simple average scoring margin in the training window. More → |
STRICT
4g
|
0.686 | 64.3% | 0.223 | 0.637 | 756 | 0.667 | 75.0% | 0.231 | 4 |
| 5 | Efficiency Efficiency Tempo-adjusted efficiency version of Pythagorean ratings. More → |
FULL
no 7d
|
0.518 | 53.6% | 0.350 | 1.121 | 661 | - | - | - | 0 |
| 6 | Margin Margin Linear team-strength model fit on point differential instead of binary wins. More → |
STRICT
4g
|
0.474 | 47.9% | 0.286 | 0.775 | 756 | 0.333 | 50.0% | 0.249 | 4 |
| 7 | Points Off/Def Points Off/Def Raw points regression with separate offensive and defensive team parameters. More → |
STRICT
4g
|
0.521 | 52.2% | 0.291 | 0.802 | 756 | 0.333 | 50.0% | 0.249 | 4 |
| 8 | Home Team Baseline Home Team Baseline Always favor the home team with a fixed prior. More → |
STRICT
4g
|
0.539 | 54.0% | 0.252 | 0.697 | 756 | 0.167 | 25.0% | 0.310 | 4 |
| 9 | Elo Elo Streaming paired-comparison rating with recency baked into sequential updates. More → |
STRICT
4g
|
0.544 | 52.4% | 0.252 | 0.698 | 756 | 0.000 | 0.0% | 0.375 | 4 |
| 10 | Bradley-Terry Bradley-Terry Static logistic paired-comparison model with one team strength parameter. More → |
STRICT
4g
|
0.486 | 48.4% | 0.273 | 0.745 | 756 | 0.000 | 25.0% | 0.272 | 4 |
| 11 | Bradley-Terry Recency Bradley-Terry Recency Static Bradley-Terry with exponential recency weights on newer games. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 0.0% | 0.367 | 4 |
| 12 | Margin Recency Margin Recency Margin regression with exponential recency weights on newer games. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 25.0% | 0.296 | 4 |
| 13 | Points Off/Def Recency Points Off/Def Recency Off/def points regression with exponential recency weights. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 25.0% | 0.293 | 4 |
| 14 | Core Ensemble Core Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and points off/def. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 25.0% | 0.299 | 4 |
| 15 | Recency Ensemble Recency Ensemble Equal-logit blend of Elo, recency BT, recency margin, log-adjusted pyth, and recency points off/def. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 25.0% | 0.309 | 4 |
| 16 | Dynamic Bradley-Terry Dynamic Bradley-Terry Time-evolving paired-comparison model with latent team strength drift. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 0.0% | 0.332 | 4 |
| 17 | Adjusted Context Blend Adjusted Context Blend Experimental context-heavy win model blending strong team components with rest and venue context. More → |
STRICT
4g
|
- | - | - | - | 0 | 0.000 | 25.0% | 0.299 | 4 |
Methodology
ELO / Bradley-Terry
- ELO: Iterative updates, K=64, HCA=100
- BT: Static logistic regression on all games
- Both model win probability, not margin
- ELO updates after each game; BT fits once
Pythagorean Models
- Raw: Classic points scored/allowed formula
- Efficiency: Pace-adjusted (pts per possession)
- Adjusted: Opponent-adjusted efficiency
- Log: Log-linear multiplicative scale
Margin Regression
- Team-level ridge regression on point margin
- Linear Bradley-Terry (margin target)
- Alpha=0.05 (CV-tuned)
- Learns home advantage from data (~6 pts)
Baselines
- Home Team: Always predict home wins (60%)
- Avg Margin: Higher average margin wins
- Models should beat these to add value