MODEL STATUS
Prometheus Engine
Find your threshold. Know your probability.
Our NFL probability engine is reviewed and fine-tuned weekly as new game results arrive. Improvements are tested before they reach your selections.
Weeks 1–4 · Alternate-threshold accuracy
How often players reached the selected alternate threshold, grouped by the fitted model’s probability band.
| Probability band | Actual accuracy |
|---|---|
| 20–25% | 22.6% |
| 25–30% | 27.9% |
| 30–35% | 32.5% |
| 35–40% | 41.4% |
| 40–45% | 39.2% |
| 45–50% | 50.8% |
| 50–55% | 50.8% |
| 55–60% | 54.6% |
| 60–65% | 60.9% |
| 65–70% | 67.2% |
| 70–75% | 71.0% |
| 75–80% | 74.9% |
| 80–85% | 81.3% |
| 85%+ | 84.7% |
Retrospective fitted-model results across 10,533 alternate thresholds and 56 games. Weeks 1–4 were used in fitting; these are not independent live-test results. Different thresholds from the same player or game are related. Anytime touchdowns are excluded from this historical dataset. The fitted research model is evaluated separately from the current live model.
Week 5 · Live results
Live alternate thresholds only. Pending, push and void results are excluded from accuracy. Results refresh from graded game data; weekly results remain available below.
Accuracy measures thresholds hit, not betting profit. Past performance does not guarantee future results.
Explore the Prometheus Engine →