Elo ratings
CFBD Elo is a sequential team-strength rating. Each completed game updates the two teams' pregame ratings based on the result, the difference between their ratings, the scoring margin, and home-field context for non-neutral games.
Higher values indicate a stronger rating. Elo is useful when an analysis needs a strength estimate that evolves one game at a time.
How it behaves
The model enters a game with a rating for each team and produces an updated rating afterward. Results that are more surprising relative to the pregame ratings generally cause larger changes, while expected results cause smaller changes. Scoring margin also affects the size of the update.
Because each update starts from the previous rating, Elo carries information forward through the sequence of games. It is therefore path-dependent: the order and history of results matter, not just a team's final record or average margin.
When an FBS team plays an opponent without an available Elo rating, the rated team's Elo remains unchanged for that game.
CFBD Elo is proprietary. The public documentation does not include its exact update formula, constants, or initialization values.
Reading the API response
The elo field is the team's latest available rating through the requested
season boundary. Higher is better, but the value is relative rather than a
direct estimate of points or win probability.
The optional week filter returns the latest completed regular-season game at or before that week. A null value means no applicable stored rating was available for the requested team and boundary.
Limits to keep in mind
Elo is sensitive to starting values and carries prior results forward. Early ratings can reflect limited current-season evidence, while later ratings are informed by a longer sequence of games.
The rating does not directly explain why a team won, separate offense from defense, or replace a matchup-specific win probability model.
Use the Elo ratings operation to filter by season, week, season type, team, or conference.