Next game
The next run happens about four days before kickoff, then again after the final injury report and on gameday.
Season outlook
Seattle is 3-3. The rest of the regular season, simulated 10,000 times as of 2026-10-19.
Share of simulations ending on each win total (a tie counts half; .5 rounds up). Green bars are inside the 80% range.
P(playoffs) from each daily simulation.
Simplified tiebreakers. Games simulated independently; team strength is not resampled.
This week
How the win probability moved across the three runs.
Why
Contribution to the projected margin, in points.
Home-field advantage is part of the model's baseline, so it isn't listed here.
Season log
Predictions are append-only. Each game is scored on the last prediction made before kickoff.
| Kickoff | Matchup | Runs (win %) | Final pick | Result | Call |
|---|---|---|---|---|---|
| 2026-09-10 | vsNE | — | — | W13–10 | Not predicted |
| 2026-09-20 | atARI | — | — | W31–7 | Not predicted |
| 2026-09-27 | atWAS | — | — | L31–33 | Not predicted |
| 2026-10-04 | vsLAC | — | — | W30–23 | Not predicted |
| 2026-10-11 | vsSF | 46›50›50 | 50% | L29–32 | Missed |
| 2026-10-16 | atDEN | 47›53›55›63 | 63% | L14–17 | Missed |
| 2026-10-26 | vsKC | — | — | ·Scheduled | Scheduled |
| 2026-11-03 | vsCHI | — | — | ·Scheduled | Scheduled |
| 2026-11-08 | vsARI | — | — | ·Scheduled | Scheduled |
| 2026-11-15 | atLV | — | — | ·Scheduled | Scheduled |
| 2026-11-29 | atSF | — | — | ·Scheduled | Scheduled |
| 2026-12-08 | vsDAL | — | — | ·Scheduled | Scheduled |
| 2026-12-13 | vsNYG | — | — | ·Scheduled | Scheduled |
| 2026-12-19 | atPHI | — | — | ·Scheduled | Scheduled |
| 2026-12-26 | vsLA | — | — | ·Scheduled | Scheduled |
| 2027-01-03 | atCAR | — | — | ·Scheduled | Scheduled |
| 2027-01-10 | atLA | — | — | ·Scheduled | Scheduled |
Live this season
Predictions are logged ~1 day before kickoff and never edited. Lower log-loss is better; a coin flip scores 0.693.
| Method | Log-loss | Brier | Accuracy | Games |
|---|---|---|---|---|
| This model | 0.730 | 0.254 | 51.9% | 28 |
| Vegas spread Benchmark | 0.687 | 0.235 | 63.0% | 28 |
| Elo only Baseline | 0.697 | 0.241 | 55.6% | 28 |
Compared on the 28 completed games that have a Vegas line. Ties count as half a win.
Model report
Every season is predicted by a model trained only on earlier seasons. Lower log-loss is better; a coin flip scores 0.693.
| Method | Log-loss | 90% CI | Brier | Accuracy |
|---|---|---|---|---|
| This model | 0.626 | 0.617–0.634 | 0.218 | 65.0% |
| Elo only Baseline | 0.633 | 0.623–0.643 | 0.220 | 64.4% |
| Home team always Baseline | 0.686 | 0.682–0.690 | 0.245 | 56.1% |
| Vegas spread Benchmark | 0.613 | 0.604–0.622 | 0.212 | 66.1% |
Predicted vs actual home-win rate per probability bin. Dot size = games.
Model log-loss in each backtest season.