Running record vs. the baselines
Model results cover every issued pick with a final score. Each baseline covers only games with valid closing odds; missing odds are shown separately. Baselines are graded on the closing line from the nflverse schedules file. Brier score is the average squared error of the published probability; lower is better, 0.25 is a coin flip.
All graded picks (15 games)
| Who | W | L | Push | Missing | Win % | Brier |
| Gridiron model | 10 | 4 | 1 | 0 |
71.4% |
0.2186 |
| Always home | 5 | 9 | 1 | 0 |
35.7% |
|
| Closing-spread favorite | 9 | 5 | 1 | 0 |
64.3% |
0.2275 |
| Moneyline favorite | 9 | 5 | 1 | 0 |
64.3% |
0.2267 |
2026 season
| Who | W | L | Push | Missing | Win % | Brier |
| Gridiron model | 10 | 4 | 1 | 0 |
71.4% |
0.2186 |
| Always home | 5 | 9 | 1 | 0 |
35.7% |
|
| Closing-spread favorite | 9 | 5 | 1 | 0 |
64.3% |
0.2275 |
| Moneyline favorite | 9 | 5 | 1 | 0 |
64.3% |
0.2267 |
Week by week
| Season | Week | Model | Home | Spread fav | ML fav |
| 2026 | 5 | 10-4 | 5-9 | 9-5 | 9-5 |
Recent graded picks
| Week | Game | Pick | Prob | Result |
| 2026 wk 5 | TB 24 at DAL 16 | Dallas | 64% |
miss |
| 2026 wk 5 | PHI 17 at JAX 19 | Jacksonville | 80% |
hit |
| 2026 wk 5 | CHI 7 at GB 31 | Green Bay | 52% |
hit |
| 2026 wk 5 | CIN 31 at MIA 24 | Cincinnati | 68% |
hit |
| 2026 wk 5 | CLE 12 at NYJ 17 | New York (NYJ) | 51% |
hit |
| 2026 wk 5 | HOU 26 at TEN 3 | Houston | 64% |
hit |
| 2026 wk 5 | IND 22 at PIT 34 | Pittsburgh | 50% |
hit |
| 2026 wk 5 | LV 17 at NE 45 | New England | 63% |
hit |
| 2026 wk 5 | MIN 23 at NO 13 | Minnesota | 57% |
hit |
| 2026 wk 5 | NYG 27 at WAS 27 | Washington | 65% |
push |
| 2026 wk 5 | DEN 23 at LAC 17 | Los Angeles (LAC) | 52% |
miss |
| 2026 wk 5 | DET 45 at ARI 21 | Detroit | 64% |
hit |
| 2026 wk 5 | SF 32 at SEA 29 | San Francisco | 62% |
hit |
| 2026 wk 5 | BAL 30 at ATL 23 | Atlanta | 60% |
miss |
| 2026 wk 5 | BUF 37 at LA 34 | Los Angeles (LA) | 65% |
miss |
The full pick log, with pre-kickoff timestamps, is picks.csv. The record as data is record.json.