NCAA Football prediction quality
Season 2026. 26 eligible official forecasts; 23 finalized. Updated 2026-10-20T00:04:35.378Z.
Only original official forecasts with publication and creation times before kickoff qualify. Early previews, missing timestamps, and duplicate game records are excluded. The dashboard recomputes market outcomes from saved sides, lines and final scores without editing the original record. This is the site’s recorded history, not an independently certified audit.
Winner
16 wins · 7 losses · 0 pushes · 0 passes · 3 pending.
Win rate 69.6%. 95% Wilson interval 49.1%–84.4%.
Small sample. These results do not establish a betting advantage.
Confidence accuracy
| Confidence band | Picks | Average forecast | Actual win rate | 95% interval |
|---|---|---|---|---|
| 50–60% upper excluded | 5 | 56.8% | 40.0% | 11.8%–76.9% |
| 60–70% upper excluded | 9 | 64.3% | 55.6% | 26.7%–81.1% |
| 70–80% upper excluded | 4 | 74.5% | 100.0% | 51.0%–100.0% |
| 80–90% upper excluded | 4 | 83.0% | 100.0% | 51.0%–100.0% |
| 90–100% inclusive | 1 | 90.0% | 100.0% | 20.7%–100.0% |
Brier score 0.172 (lower is better); weighted absolute calibration gap 14.9%. 23 scored probabilities; 0 decided picks missing a valid probability.
Spread / ATS
12 wins · 11 losses · 0 pushes · 0 passes · 3 pending.
Win rate 52.2%. 95% Wilson interval 33.0%–70.8%.
Small sample. These results do not establish a betting advantage.
Confidence accuracy
| Confidence band | Picks | Average forecast | Actual win rate | 95% interval |
|---|---|---|---|---|
| 50–60% upper excluded | 15 | 54.6% | 46.7% | 24.8%–69.9% |
| 60–70% upper excluded | 5 | 65.8% | 60.0% | 23.1%–88.2% |
| 70–80% upper excluded | 2 | 71.8% | 50.0% | 9.5%–90.5% |
| 90–100% inclusive | 1 | 95.0% | 100.0% | 20.7%–100.0% |
Brier score 0.252 (lower is better); weighted absolute calibration gap 8.5%. 23 scored probabilities; 0 decided picks missing a valid probability.
Total
10 wins · 13 losses · 0 pushes · 0 passes · 3 pending.
Win rate 43.5%. 95% Wilson interval 25.6%–63.2%.
Small sample. These results do not establish a betting advantage.
Confidence accuracy
| Confidence band | Picks | Average forecast | Actual win rate | 95% interval |
|---|---|---|---|---|
| 50–60% upper excluded | 16 | 53.7% | 43.8% | 23.1%–66.8% |
| 60–70% upper excluded | 6 | 61.9% | 50.0% | 18.8%–81.2% |
| 70–80% upper excluded | 1 | 70.6% | 0.0% | 0.0%–79.3% |
Brier score 0.273 (lower is better); weighted absolute calibration gap 13.1%. 23 scored probabilities; 0 decided picks missing a valid probability.
Closing-line comparison
The last qualifying observation within 15 minutes of kickoff is a proxy for the closing line, not a verified final sportsbook quote. Both the saved line and later observation require receipt and provider quote timestamps, the same book/source, and matching kickoff. Unknown, stale, post-kickoff and modeled quotes are excluded. No historical quotes are reconstructed. Positive points mean a more favorable saved number for the selected side; points are not profit or ROI. Moneyline price comparisons are not included.
Spread: 0 compared / 23 eligible past picks. Average unavailable points; 0 positive, 0 negative, 0 unchanged.
Total: 0 compared / 23 eligible past picks. Average unavailable points; 0 positive, 0 negative, 0 unchanged.
0 observations archived; 0 lack receipt/quote timestamps.
Model versions
| Saved version | Market | Wins | Losses | Pushes |
|---|---|---|---|---|
| 8.0-adjusted-scoring | winner | 16 | 7 | 0 |
| 8.0-adjusted-scoring | spread | 12 | 11 | 0 |
| 8.0-adjusted-scoring | total | 10 | 13 | 0 |
Winner calibration uses displayed confidence. Football spread/total calibration uses saved selected-side probabilities; basketball uses saved raw selected-side spread/total probabilities. Pushes, passes and pending games are excluded from calibration. Wilson intervals describe binomial uncertainty and do not account for correlations among picks. Confidence estimates are diagnostic; this release does not automatically tune the prediction model.