Track record
Adjusted per-possession efficiency ratings (offense/defense PPP + tempo, opponent-adjusted at prediction time; fitted team_alpha=0.10, carryover=0.6, home edge 5.04 points). The closing line is never an input — only the benchmark. Walk-forward: every prediction made before learning the result.
Model vs the closing line
The walk-forward verdict, in full: the model is a competent forecaster and loses to the closing line everywhere that matters.
| Metric | Us | Closing line | Verdict |
|---|---|---|---|
| Margin RMSE | 12.17 | 11.04 | +1.13 worse |
| Total RMSE | 18.05 | 16.90 | +1.15 worse |
| ATS (≥0.5pt disagreement) | 16,744-17,090-473 · 50% | — | needs 52.4% at -110 |
| O/U (≥1pt disagreement) | 14,569-14,252-193 · 51% | — | negative |
ATS by era
No era comes close to a betting edge — and none of them reaches the 52.4% break-even line. The dashed line is 52.4% — break-even against standard -110 vig.
| Era | ATS % | O/U % | Straight-up | Our RMSE | Line RMSE |
|---|---|---|---|---|---|
| 2013–2016 | 49% | 52% | 70% | 11.66 | 10.52 |
| 2017–2020 | 50% | 50% | 71% | 12.13 | 11.17 |
| 2021–2026 | 50% | 50% | 71% | 12.68 | 11.35 |
52.4% is break-even against standard -110 vig. Green = profitable, red = not.
Season by season
Per-season detail is still being tabulated from the walk-forward run — the era summary below comes from the completed backtest (2013–2021, 2025–2026; seasons 2022–2024 missing after the CBBD monthly quota was exhausted mid-pull).