College Basketball Elo Ratings

Each team has an offensive rating and a defensive rating. Both increase when a team performs better than expected: offense for scoring more, defense for allowing fewer points. Overall Elo is the average of these two ratings and is used for the overall-rank summary when a team is selected.

Expected scores compare each offense with the opposing defense, using a trailing league scoring average. Home-court advantage varies by team using prior-season home and away results adjusted for opponent quality; neutral-site games receive no home boost. A logistic calibration converts expected margin into win probability.

Completed games show ratings immediately before and after that game, with a pregame probability reconstructed from the historical model. The probability calibration is trained only on seasons before the displayed season. These are model reconstructions, not an archive of forecasts published before tip-off. Source corrections or parameter changes can revise them.

Prediction accuracy counts completed, non-tied games with an available probability. A home win is predicted when the home probability is at least 50%; otherwise the away team is predicted. The selected-season figure includes games played so far, while all-time accuracy combines every eligible completed game in the available seasons.

Unplayed games all use the latest available ratings; future results are not simulated. A game awaiting a final result can appear as unplayed even after its listed tip-off. Expected scores are averages, not predicted exact scores. Probabilities are model estimates, not guarantees.

Coverage includes NCAA Division I men's basketball matchups. Exhibition games are excluded, so displayed records can differ from official records. Conference tournament and NCAA tournament matchups appear once the schedule provider lists both participants.

Model details and parameter values · Source: hoopR

The weekly charts include completed games before the next listed week's first tip-off. Weekly changes compare with the start of that window; season-to-date changes compare with preseason ratings after the offseason reset. “All weeks” shows the latest ratings. The change lists show up to 10 gains and 10 losses, excluding teams with no change. In the first available season, win probabilities are unavailable because no earlier training data exists.