1 · Collect verified results
Only games locked as FINAL up to the active week feed the model. Records, points for and against, and per-game averages are all derived from those completed games — nothing is manually entered.
Data & Ranking Methodology
How every CFB 360 rating and projection is built
No polls, no media narratives, no brand-name favoritism. Every team runs through the same repeatable framework, and this page documents exactly what goes into it.
Model inputs
Each input is measured per play or per game, then scaled against the schedule a team has actually faced.
Opponent-adjusted EPA per play on offense and defense
Down-to-down success rate and explosive-play value
Strength of schedule scaling from every completed game
Net point differential with capped margins
Roster talent composite as a decaying preseason prior
The weekly pipeline
2 · Adjust for opponent quality
Every efficiency input is weighted by the quality of the opponent it was produced against, so a dominant performance against a strong defense outranks the same numbers against a weak one.
3 · Cap blowouts and garbage time
Margin of victory is capped at 24 points for Power 4 teams and 21 for non-Power 4 teams, and fourth-quarter efficiency is filtered once the margin passes 28 points. Running up the score does not inflate a rating.
4 · Absorb change gradually
A team's rating can move at most ±3.0 points in a single week, and outlier games decay over time. One unusual result informs the model without defining a season.
Portal continuity multiplier
Multi-year history only describes a roster that still exists. Every team gets a roster continuity score built from its transfer portal traffic — additions and departures, their composite ratings, and whether the head coach is new. High turnover scales the weight of the 2021–2025 archive down and the incoming talent composite up, so a team that rebuilt through the portal is judged on the players it added rather than the seasons it no longer resembles. Continuity never falls below 15%, so the archive always anchors the baseline, and a coaching change alone counts as meaningful turnover.
Game-by-game time decay
The balance between the preseason baseline and current-season results shifts on a fixed schedule, published on every matchup page so you can see which phase a projection is in.
The baseline is the multi-year archive weighted by roster continuity, blended with the portal and recruiting talent composite for the share of the roster that turned over.
Current-season results phase in on a fixed ramp: 25% after one game, 50% after two, 75% after three, with the preseason baseline carrying the remainder.
From the fourth game on, 85%+ of the projection comes from actual 2026 performance and the historical baseline acts only as a minor anchor.
EPA adjustments
Expected points added measures what each play was actually worth in scoring terms rather than raw yardage. Offensive and defensive EPA per play are opponent-adjusted, paired with success rate, explosiveness, havoc rate and drive scoring efficiency, and drawn from the stored official archive for past seasons plus verified 2026 results as they are graded. Where a measurement has not been published, the page says so — no value is estimated to fill a gap.
Against-the-spread tracking
Every projection is graded after the fact against the closing market line and the final score. Only games with both a published line and a final result are graded, so the record shown on a matchup's betting trends is a real sample rather than a back-fitted one. Those graded results are what set each matchup's confidence tier below.
Confidence levels
Every matchup card carries a confidence rating based on how accurate the model has historically been for that category of game.
Projected margin inside 27 points, non-conference home favorites, and non-Power 4 road underdogs — the categories with the strongest historical accuracy.
Solid historical accuracy with some variance, or a projected margin sitting near the edge of the model's most reliable range.
Wide gaps between market lines and the model, blowout spreads beyond 27 points, high-edge SEC plays, and ACC road underdogs.