Methodology · College Football

How the College Football model works

Who makes the 12-team Playoff, who's on the bubble, and who wins it all.

Version
v1 (points-scale, MOV-aware)
Updated
Weekly in season
Data source
CFBD (ratings, returning production) + betting-market win totals & spreads
Calibration
Backtested on prior seasons; in-season results graded on the Ledger

The question

Every team's odds to reach the 12-team College Football Playoff, its expected win total, and its path to the national title — plus who sits just outside the field.

The data

Two signals: the betting market's win totals and game lines (a sharp, continuously-updated read on team strength) and CFBD's efficiency + returning-production data. Refreshed weekly during the season.

How strength is measured

Teams are rated on a points scale — a market-anchored rating blended with a season-outlook layer built from efficiency and roster continuity. That blended rating is the public board order.

Recency & regression

In-season, ratings update toward results with a margin-of-victory-aware adjustment (blowouts move the needle more than one-score games, with diminishing returns). Preseason, teams are regressed toward priors because early-season samples are tiny.

Injuries & roster

Most injury news is already priced by the market — lines move, and we follow. Material news (a starting QB out for the year) enters as a documented point adjustment with a reason and a date, never a hidden thumb on the scale.

Home advantage

A standard home-field bump is applied on the points curve when converting the rating gap into a single-game win probability.

The simulation

A Monte Carlo runs the full schedule thousands of times, producing each team's win distribution, then builds the 12-team bracket: five conference-champion auto-bids plus seven at-large teams, top four seeded on a bye.

Postseason format

The 12-team CFP: seeds 1–4 get a first-round bye; 5–12 play the opening round on campus. Access protection is applied where the format requires it.

Backtesting

The win curve is calibrated so that teams given a 70% game win probability win about 70% of the time. From 2026 on, weekly forecasts are timestamped to the Ledger and graded against results.

Known weaknesses

Preseason numbers carry real uncertainty; the market can be noisy week to week; and the model reads team strength, not the specific matchup chaos that makes college football college football.

Every published number is timestamped to the Ledger the moment it's made and graded against results — nothing here is backdated. Percentages are rounded for readability.