2025-2026 NBA Model Performance Analysis
Comparing prediction accuracy across 1305 games using multiple rating models.
Model Performance Leaderboard
Models ranked by AUC on 2025-2026 games. Hover over column headers for explanations.
| # | Model | AUC | Acc | Brier | LogLoss | n | AUC 7d | Acc 7d | Brier 7d | n 7d |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Avg Margin (Baseline) | 0.757 | 69.1% | 0.203 | 0.593 | 1177 | 0.500 | 33.3% | 0.252 | 3 |
| 2 | Pythagorean Efficiency | 0.749 | 67.6% | 0.204 | 0.594 | 1177 | - | - | - | 0 |
| 3 | Points O/D | 0.748 | 67.2% | 0.209 | 0.607 | 1177 | 0.250 | 33.3% | 0.268 | 3 |
| 4 | Pythagorean Raw | 0.748 | 68.1% | 0.216 | 0.622 | 1177 | 0.500 | 33.3% | 0.250 | 3 |
| 5 | Pythagorean Adjusted | 0.746 | 67.6% | 0.205 | 0.597 | 1177 | 0.500 | 33.3% | 0.255 | 3 |
| 6 | Pythagorean Log | 0.746 | 67.6% | 0.205 | 0.597 | 1177 | 0.500 | 33.3% | 0.255 | 3 |
| 7 | Bradley-Terry | 0.744 | 67.8% | 0.207 | 0.601 | 1177 | 0.500 | 33.3% | 0.282 | 3 |
| 8 | Margin Regression | 0.739 | 67.3% | 0.212 | 0.613 | 1177 | 0.250 | 33.3% | 0.267 | 3 |
| 9 | Pure ELO | 0.738 | 68.1% | 0.208 | 0.604 | 1177 | 0.500 | 66.7% | 0.223 | 3 |
| 10 | Home Team (Baseline) | 0.549 | 54.9% | 0.250 | 0.694 | 1177 | 0.250 | 33.3% | 0.293 | 3 |
| 11 | Bradley-Terry Recency | - | - | - | - | 0 | 0.500 | 66.7% | 0.225 | 3 |
| 12 | Margin Recency | - | - | - | - | 0 | 0.250 | 66.7% | 0.249 | 3 |
| 13 | Points O/D Recency | - | - | - | - | 0 | 0.250 | 66.7% | 0.250 | 3 |
| 14 | Core Ensemble | - | - | - | - | 0 | 0.250 | 66.7% | 0.238 | 3 |
| 15 | Recency Ensemble | - | - | - | - | 0 | 0.250 | 66.7% | 0.235 | 3 |
| 16 | Adjusted Context Blend | - | - | - | - | 0 | 0.250 | 33.3% | 0.264 | 3 |
| 17 | Dynamic Bradley-Terry | - | - | - | - | 0 | - | 66.7% | 0.245 | 3 |
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: Bill James formula (pts scored/allowed)
- Efficiency: Pace-adjusted (pts per possession)
- Adjusted: KenPom-style opponent adjustment
- 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