How good are these forecasts?
Every forecast scored on seasons the model never saw while its settings were chosen.
Held-back seasons
| Forecast | Forecasts | Log loss | Favourite won |
|---|---|---|---|
| The model | 13235 | 0.503 | 74% |
| Base rate | 13235 | 0.629 | 68% |
| Recent form | 13235 | 0.557 | 71% |
Against the betting markets
Each market's price before the start, scored on the same finished games as the model. A negative figure means the market was more accurate.
| Market | Since | Priced | Model log loss | Market log loss | Model vs market |
|---|---|---|---|---|---|
| Bookmakers (Pinnacle, DraftKings, bet365) | 19 Jan 2026 | 2619 | 0.570 | 0.545 | −4.6% |
| Polymarket | 8 Feb 2025 | 6599 | 0.501 | 0.483 | −3.6% |
| Kalshi | 3 Nov 2025 | 5588 | 0.516 | 0.483 | −6.9% |
Every Division I game since 2003 moves the ratings, and so do games against teams from lower divisions; a team new to the file starts 150 points below average. Only Division I teams are in the tables and the title odds.
The title odds use an approximate tournament: each conference's best team by record gets in, the rest of the 64 are the best left by simulated wins and rating, seeded on an S-curve on neutral floors. The committee, conference tournaments and the First Four are not modelled.
Roster turnover between seasons is a pull toward average in the rating; the adjustment for who is playing adds what each returning player and transfer did in their box scores, but freshmen start at a bench player's value and recruiting rankings are not used.
Between April and November there are no games; the pages show how the last season finished.
Questions and answers
How accurate are PodiumOracle's NCAA men's basketball predictions?
On 13,235 games from held-back seasons (2024–25 to 2026–27), which the model never saw while it was tuned, the side it favoured won 74% of the time. Its log loss was 0.503, against 0.629 for the base rate and 0.557 for recent form. Against the betting markets on the same finished games, its log loss was 0.570 against the bookmakers' 0.545 over 2,619 games, 0.501 against Polymarket's 0.483 over 6,599 games and 0.516 against Kalshi's 0.483 over 5,588 games (lower is better).