2026 MODEL PERFORMANCE
How is the model performing?
Every number below comes from frozen pregame PDR recommendations. Published picks stay frozen; this page simply scores what happened next.
HOW TO READ THIS PAGE
SettledOnly WIN and LOSS decisions count toward hit rate. Voids and pending plays stay out.
Price-coveredROI uses only settled plays where the frozen sportsbook price was stored at publication.
CalibrationCompares stated model confidence with observed outcomes. It measures probability accuracy, not just pick record.
ROI coverage is incomplete.
Frozen side prices exist for 816 of 937 settled plays (87.1% coverage); 121 settled plays are excluded from units and ROI.
Current price-covered history begins after the earliest published results, so ROI should not be treated as the season's complete betting return.
OBSERVED RESULTS
Where the results are coming from
These splits are descriptive and can move as the sample grows.
By week
All frozen published recommendations.
| Week | Record | Hit | Pending |
|---|---|---|---|
| Week 1 | 60-61-0 (7 voids) | 49.6% | 0 |
| Week 2 | 54-51-0 (1 void) | 51.4% | 0 |
| Week 3 | 63-64-0 (2 voids) | 49.6% | 0 |
| Week 4 | 84-50-0 (6 voids) | 62.7% | 0 |
| Week 5 | 63-55-0 | 53.4% | 0 |
| Week 6 | 43-43-0 (3 voids) | 50.0% | 0 |
| Week 7 | 59-52-0 | 53.2% | 0 |
| Week 8 | 78-57-0 (1 void) | 57.8% | 12 |
OVER vs UNDER
Side-level results across the current season.
| Side | Plays | Record | Hit |
|---|---|---|---|
| OVER | 401 | 209-175-0 (10 voids) | 54.4% |
| UNDER | 568 | 295-258-0 (10 voids) | 53.3% |
By market
Market splits are descriptive, not proof of a persistent edge.
| Market | Plays | Record | Hit |
|---|---|---|---|
| Receiving yards | 311 | 163-134-0 (9 voids) | 54.9% |
| Receptions | 298 | 164-125-0 (7 voids) | 56.7% |
| Rushing yards | 162 | 75-80-0 (4 voids) | 48.4% |
| Passing yards | 125 | 60-65-0 | 48.0% |
| Passing TDs | 73 | 42-29-0 | 59.2% |
Probability calibration: confidence vs. outcomes
This compares the model's average stated win probability with the observed win rate in settled WIN/LOSS decisions. Smaller gaps mean better calibration.
| Model bucket | N | Avg model | Observed | |
|---|---|---|---|---|
| 57.5%-59.9% | 85 | 58.8% | 51.8% | |
| 60.0%-64.9% | 155 | 62.9% | 51.6% | |
| 65.0%-69.9% | 214 | 67.6% | 53.3% | |
| 70.0%-74.9% | 190 | 72.4% | 57.9% | |
| 75.0%+ | 293 | 81.0% | 53.2% |
Read this as an early-season scorecard, not a guarantee.
The prospective sample is still developing. Calibration, hit rate, and ROI can move materially as more frozen recommendations settle.
PDR does not retroactively change published picks.