🏈 CFB Modelopponent-adjusted scoring matrix · 2016–2024
—
pick teams

Head-to-head scoring matrix

Each factor's net edge (home − away), standardized, then scaled to points by the model's calibration. Positive = favors home.

Factor weights

Tune the matrix, or auto-search the weights. Changes re-project instantly and can be re-backtested.

Walk-forward: each season graded using only prior seasons.

Predicted vs actual margin

Each dot is a game (home perspective). Tight diagonal = accurate.

Edge vs ATS outcome

Does a bigger model edge convert to covers?

ATS win rate by season

Cumulative units (flat −110)

Bankroll curve betting every model ATS pick.

🐕 Rushing Dogs — the validated edge

Small underdogs (≤7) whose run game beats the favorite's. Backtest: 56.9% ATS over 833 games (2016–24), +8.6% ROI, every season above break-even, holds out-of-sample. Bet at the close. Home dogs = higher conviction (58.8%), road dogs = stable core (55.9%). Ratings use last season's stats early, blending to current by ~game 4.