NCAA Football — Spread
Methodology
Picks qualify at edge ≥ 6 pts. Spread capped at ±23 for the paid tiers; picks inside ±14 are marked Premium ★ Best Bet. The outer cap was 14 until 2026-08-30, which silently discarded 150 graded picks that returned +12.1% — the tighter band has the better RATE and the worse TOTAL return. 23 is a product choice inside a flat region, not a derived optimum: 20, 23 and no cap are statistically indistinguishable (z = 0.50). Walk-forward validated; 2026 season opens August. Phase 0d corrected the non-event predicate to keep legitimate home shutouts.
All figures computed from logged pick data at build · updated with each deploy.
Per-band breakdown
Each band shows sample size, record, win rate, and the 95% Wilson confidence interval. Bands inside the production gate are shown — wider edges have higher conviction but smaller samples.
| Edge band | n | W–L | Win % | 95% Wilson CI |
|---|---|---|---|---|
| 6–8 pts | 97 | 61–36 | 62.9% | [53.0, 71.8] |
| 8–10 pts | 82 | 51–31 | 62.2% | [51.4, 71.9] |
| 10–14 pts | 128 | 80–48 | 62.5% | [53.9, 70.4] |
| 14+ pts | 147 | 94–53 | 63.9% | [55.9, 71.3] |
| Total | 454 | 286–168 | 63.0% | [58.5, 67.3] |
BBMI vs Vegas
Mean absolute error in predicting the home cover margin. Lower is a better predictor; Vegas's lower MAE means the market is the tighter predictor on this product.
n = 889 graded games