We built this to be defensible, not to impress. Every number on the site comes from standard, named football analytics computed on real data — never an invented formula. Here's exactly how it works, and what it can and can't do.
Eleven seasons of college football (2014–2024) from CollegeFootballData: every play, drive, game, and box score, plus ratings (SP+, FPI, SRS, Elo), betting lines, weather, travel, talent and returning production. Roughly 8,300 FBS games.
Raw stats lie — 6 yards a play against a bad defense isn't 6 yards against a good one. We use EPA (expected points added) and success rate as the efficiency base, opponent-adjust them via ridge regression as of each week (so a Week 8 number only knows Weeks 1–7), and strip garbage time. Player value uses WEPA— the same EPA, opponent-adjusted — and a PAAR-style value over a replacement-level backup. These are the field's standards, not ours.
Game predictions blend two of the strongest public power ratings via a formula validated out-of-sample over 8,300 FBS games (2014–2024):
margin = 0.45 × Elo_margin + 0.55 × SP+_margin + 2.4 (home field)
Elo_margin = (homeElo − awayElo) ÷ 25
SP+_margin = homeSP+ − awaySP+
win_prob = normCDF(margin ÷ 13.5)
Elo (45%) tracks strength through results — it rises when you beat good teams, falls when you lose to bad ones, and resets each season. It captures momentum and form. SP+ (55%) tracks per-play efficiency independent of score and tempo — pass-success rate, rushing explosiveness, third-down conversion, all opponent-adjusted. It catches teams that win ugly and teams that lose close. The blend (45/55) was chosen by walk-forward grid search; SP+ earns the higher weight because efficiency predicts future results better than past results do, especially early in the season.
The 13.5-point sigma converts a projected margin to a win probability via a normal distribution — this matches CFB final-margin variance, validated on 2014–2024 OOS data. When you tap the model pick badge on any game in Pick'em, you can see both components and the edge vs. the market line.
Everything is walk-forward backtested: the model is only ever scored on games it was trained before, never on data it has seen. We caught our own inflated early numbers (a closing-line leak), retracted them, and report only what survives a leak-free, out-of-sample test:
Closing Line Value is the metric a sharp book respects: our number tends to land where the market closes, which is the real, measurable sign of an edge — not a fabricated win rate.
We are nota Vegas-beater. The closing line is the sharpest number in sports and nobody reliably beats it; anyone selling you 70%+ winners is lying. If one of our backtests prints 58%+ ATS, that's leakage, and we'd rather tell you that than sell it. We report losses next to wins. We never call an EPA-based value a “grade” (we can't source charted grades, snaps, routes or air yards), and we label usage as play share, not snaps.
For player availabilitywe serve our own stored copy of the feed and ingest the official SEC / Big Ten / ACC / Big 12 / CFP availability reports as they publish each game week (ESPN's aggregation in the meantime). We treat it as context, not a betting signal — the market prices known absences efficiently, so we show it to inform you, not to sell you an edge.
Questions about the method, or want to see a number defended? Read how Gridpex works and makes money.