Picks vs outcomes
What each model predicted, and what actually happened, graded against official results. Paper-only and educational — nothing here is a pick or a recommendation to wager, and no stake is filled in anywhere on this site.
| Sport | Graded | Hit | Missed | Rate | What is being graded |
|---|---|---|---|---|---|
| MLB | 38,519 | 19,312 | 19,207 | 50.1% | Player-prop projections: the model's over/under lean against the book's line, graded from the official box score. |
| NFL | 43 | 24 | 19 | 55.8% | Preseason game forecasts: which side the model favoured, graded against the official final. |
| UFC | 16 | 10 | 6 | too few to say | Fight-winner picks: the model's chosen fighter and its probability, graded against the official result. |
| Premier League | 24 | 11 | 13 | too few to say | Match-result forecasts: the outcome the model gave the most probability to, graded against the official full-time score. |
These four numbers are not comparable with each other. They come from different models answering different questions over samples that differ by four orders of magnitude — thousands of player-prop projections against a handful of fight winners — so reading down this column and picking a best model is not something the numbers support. Each row links to its own record, where what it covers is stated in full.
A hit rate is also not a claim about a sportsbook. Being right more often than not and being right more often than the posted price implies are different things, and only the second would say anything about a market. That comparison is tracked separately for each sport and has cleared nothing here.
MLB · game predictions — a separate record
Winner, total and run line for each game, graded from the newest prediction revision generated before first pitch: moneyline 254–251 · total 240–242–22 · run line 319–186. Not comparable with the player-prop record above and never combined with it — full game-level record.