Model Performance
Frozen walk-forward backtest, 2016–2025
Every prediction below was made walking forward through each season with only the data available at the time, then frozen. The market benchmark is the closing spread: the strongest public forecast of a game's margin. Beating it consistently is rare, and the model is measured against it, not against a naive baseline. The published model line blends the pure model with the market at a capped weight, so the two are correlated by construction; the pure model's own error runs about half a point higher.
Games graded2761
Model MAE9.99points
Market MAE9.78points
Gap to market+0.21points (lower is better)
Model closer45.6%
| Season | Games | Model MAE | Market MAE | Gap | Model closer | Bias |
|---|---|---|---|---|---|---|
| 2016 | 267 | 9.03 | 9.04 | -0.01 | 48.7% | -1.03 |
| 2017 | 267 | 10.28 | 10.08 | +0.20 | 47.2% | -0.53 |
| 2018 | 267 | 10.17 | 9.87 | +0.30 | 47.9% | +0.10 |
| 2019 | 267 | 10.42 | 10.18 | +0.24 | 43.1% | +1.68 |
| 2020 | 269 | 9.92 | 9.79 | +0.13 | 46.5% | +0.94 |
| 2021 | 285 | 11.14 | 10.67 | +0.47 | 36.1% | -0.16 |
| 2022 | 284 | 9.01 | 8.78 | +0.22 | 47.2% | -0.41 |
| 2023 | 285 | 10.38 | 9.98 | +0.39 | 42.1% | -1.03 |
| 2024 | 285 | 9.85 | 9.70 | +0.15 | 47.0% | -0.51 |
| 2025 | 285 | 9.66 | 9.67 | -0.01 | 50.5% | -0.53 |
| All seasons | 2761 | 9.99 | 9.78 | +0.21 | 45.6% | -0.16 |
Bias is the mean signed error of the model's home margin: positive means the model leans toward home teams. Seasons 2016–2021 tuned the model's parameters; 2022 onward was never touched by selection and is the honest read on accuracy.