Performance
Published picks are measured at their posted price, with coverage and losses visible.
Did our posted prices beat the sportsbook's closing line? Positive cents means we published a better number than the market's final price — the hardest benchmark to fake.
Stated probabilities matched observed outcomes across 20,000 graded probability rows. The curve plots each probability bucket against how often it actually won — dots on the diagonal mean the number says what it means.
Full reliability diagramROI is graded at the immutable posted price and stays blank until a sport has price-covered picks. CLV is the sample-weighted average across that sport's markets from the closing-line section above.
| Market | Win % | ROI | Sample | Brier |
|---|---|---|---|---|
| Moneyline | 55.5% | — | 1,307 | 0.2442 |
| Spreads | 55.9% | — | 952 | 0.2484 |
| Totals | 50.3% | — | 1,131 | 0.2634 |
| Batter Walks | 62.0% | — | 5,505 | 0.2713 |
| Batter Hits | 55.6% | — | 1,918 | 0.3244 |
| Player Points | 49.6% | — | 1,283 | 0.2784 |
| Batter Home Runs | 65.9% | — | 1,160 | 0.2360 |
| Player Rebounds | 50.8% | — | 1,118 | 0.2645 |
| Player Assists | 50.1% | — | 868 | 0.2772 |
| Player Threes | 53.3% | — | 825 | 0.2634 |
| Batter RBIs | 48.7% | — | 673 | 0.3734 |
| Pitcher Strikeouts | 51.7% | — | 633 | 0.3181 |
| Pitcher Outs | 53.0% | — | 523 | 0.3240 |
| Batter Runs Scored | 49.6% | — | 482 | 0.3651 |
| Pitcher Earned Runs | 52.9% | — | 429 | 0.3219 |
| Player Goals | 92.7% | — | 424 | 0.0575 |
| Pitcher Hits Allowed | 49.3% | — | 371 | 0.3333 |
| Player Blocked Shots | 49.6% | — | 135 | 0.2631 |
| Player Blocks | 49.4% | — | 83 | 0.2562 |
| Player Steals | 51.8% | — | 56 | 0.2682 |
| Pitcher Walks | 48.8% | — | 41 | 0.3339 |
| Player Shots On Goal | 100.0% | — | 26 | 0.0023 |
Win rate and ROI are graded from the posted pick and its recorded price-at-post. Player-prop markets show ROI once we have reliable closing-price coverage; thinner markets show Brier score + sample only until coverage catches up.
- Every published pick freezes its price, book, line, model version, and probabilities before the game.
- Same-book closing-line value is reported with coverage; missing closes remain blank.
- Calibration is computed from graded probability rows and shown with its sample size.
- Corrections are appended with prior values and reasons instead of silently rewriting the ledger.