NFL: every call, graded
About this page
Last graded
Winners calledOnly picks locked before the start count. A pick made later, or priced after the fact, never enters it.
Locked before the start, graded afterThe same 106 priced games, winners only. Where a game is priced our pick is built on the market's price, so the two mostly match; we claim no edge. When we took the other side: 3–4.
Every graded pick, priced or not
Won 65. Its stated probabilities expected 67.8 (90% range 59.8–75.8): inside that range, in line with its probabilities.
By method: model alone 20–13 (33), market-anchored 45–28 (73).
Backtest, not in the record: 63.7% of 1,693 games (walk-forward backtest of the model alone, 2020–25 seasons).
Against the betting favoriteSame games, winners only. The favorite is the side with the shorter no-vig moneyline when the pick was locked.
| Games | Brier: ours | market | Log loss: ours | market |
|---|---|---|---|---|
| All scored picks n 106 | 0.232 | 0.233 | 0.658 | 0.662 |
| Model alone n 33 | 0.231 | 0.229 | 0.654 | 0.653 |
| Market-anchored n 73 | 0.232 | 0.235 | 0.659 | 0.665 |
Lower is better for both. Each row scores our stated win probability and the market's no-vig probability on the same games and outcomes.
Picks locked before Sept 26 are the model alone; since then a priced game's number is built mostly on the market's no-vig price. Under 30 games a record is too early to read either way. Every pick, one row each →
Expected vs actual winnersThe dashed line is the wins its own probabilities expected, shaded for luck's 90% band. Solid is what happened.
Dotted line, pick 33: the method changed. Before it the calls were the model alone; after it a priced game’s number follows the market’s price, so the stretches are not the same product.
The solid line ends inside the shaded band: the record is where honest probabilities would put it.
Show as a table
| After pick | Date | Won | Expected | 90% band |
|---|---|---|---|---|
| 13 | Sep 13 | 9 | 8.0 | 5.2–10.9 |
| 27 | Sep 20 | 15 | 16.9 | 12.8–21.0 |
| 40 | Sep 27 | 24 | 25.6 | 20.7–30.5 |
| 53 | Oct 4 | 31 | 33.8 | 28.2–39.5 |
| 66 | Oct 11 | 39 | 42.5 | 36.2–48.8 |
| 80 | Oct 18 | 47 | 51.1 | 44.1–58.0 |
| 93 | Oct 22 | 56 | 59.5 | 52.1–67.0 |
| 106 | Oct 26 | 65 | 67.8 | 59.8–75.8 |
When it said 70%, did it win 70%?Each lane is a confidence band. Calibrated means the dot lands in its gray band; the bar shows luck's range.
- Said 50–60%51% won 22 of 4390% range 39–63%backtest 54% of 656
- Said 60–70%67% won 22 of 3390% range 52–78%backtest 64% of 517
- Said 70–80%70% won 19 of 2790% range 55–82%backtest 71% of 360
- Said 80–90%67% won 2 of 3only 3 graded: too few to judgebacktest 84% of 142
- Said 90–100%no live picks yetbacktest 94% of 18
Live picks, locked before the start. Backtest: walk-forward backtest of the model alone, 2020–25 seasons, not in the record. Lanes with fewer than 10 graded picks are faded.
Every graded gameLatest graded games: the model's pick, its win % and whether it won. Tap one to see why.
- DAL 38 at PHI 21Mon, Oct 26, 8:15 PMDAL 61%✓ won
- KC 3 at SEA 20Sun, Oct 25, 8:20 PMSEA 60%✓ won
- LAR 35 at LV 21Sun, Oct 25, 4:25 PMLAR 61%✓ won
- GB 17 at DET 20Sun, Oct 25, 4:25 PMDET 63%✓ won
- DEN 30 at ARI 16Sun, Oct 25, 4:05 PMDEN 67%✓ won
- IND 10 at MIN 16Sun, Oct 25, 1:00 PMMIN 76%✓ won
- SF 20 at ATL 29Sun, Oct 25, 1:00 PMSF 60%✗ lost
- CLE 27 at TEN 24Sun, Oct 25, 1:00 PMCLE 56%✓ won
- NYG 27 at HOU 21Sun, Oct 25, 1:00 PMHOU 79%✗ lost
- CIN 35 at BAL 13Sun, Oct 25, 1:00 PMCIN 56%✓ won
- TB 43 at CAR 13Sun, Oct 25, 1:00 PMCAR 62%✗ lost
- MIA 28 at NYJ 25Sun, Oct 25, 1:00 PMNYJ 63%✗ lost
- PIT 16 at NO 10Sun, Oct 25, 9:30 AMPIT 60%✓ won
- NE 24 at CHI 30Thu, Oct 22, 8:15 PMCHI 58%✓ won
- WSH 17 at SF 23Mon, Oct 19, 8:15 PMSF 77%✓ won
- DAL 13 at GB 23Sun, Oct 18, 8:20 PMGB 53%✓ won
- LAC 6 at KC 22Sun, Oct 18, 4:25 PMKC 78%✓ won
- BUF 33 at LV 30Sun, Oct 18, 4:25 PMBUF 68%✓ won
- ARI 18 at LAR 16Sun, Oct 18, 4:05 PMLAR 85%✗ lost
- TEN 26 at IND 20Sun, Oct 18, 1:00 PMIND 73%✗ lost
- PIT 26 at TB 19Sun, Oct 18, 1:00 PMPIT 56%✓ won
- CAR 25 at PHI 14Sun, Oct 18, 1:00 PMCAR 56%✓ won
- NO 34 at NYG 27Sun, Oct 18, 1:00 PMNYG 54%✗ lost
- NYJ 19 at NE 26Sun, Oct 18, 1:00 PMNE 77%✓ won
- BAL 10 at CLE 13Sun, Oct 18, 1:00 PMBAL 60%✗ lost
- CHI 31 at ATL 23Sun, Oct 18, 1:00 PMCHI 50%✓ won
- HOU 34 at JAX 31Sun, Oct 18, 9:30 AMJAX 57%✗ lost
- SEA 14 at DEN 17Thu, Oct 15, 8:15 PMSEA 56%✗ lost
- BUF 37 at LAR 34Mon, Oct 12, 8:15 PMLAR 59%✗ lost
- BAL 30 at ATL 23Sun, Oct 11, 8:20 PMATL 61%✗ lost
- SF 32 at SEA 29Sun, Oct 11, 4:25 PMSEA 60%✗ lost
- DET 45 at ARI 21Sun, Oct 11, 4:25 PMDET 67%✓ won
- DEN 23 at LAC 17Sun, Oct 11, 4:05 PMDEN 63%✓ won
- HOU 26 at TEN 3Sun, Oct 11, 1:00 PMHOU 72%✓ won
- CIN 31 at MIA 24Sun, Oct 11, 1:00 PMCIN 73%✓ won
- CHI 7 at GB 31Sun, Oct 11, 1:00 PMCHI 52%✗ lost
- NYG 27 at WSH 27Sun, Oct 11, 1:00 PMWSH 61%✗ lost
- IND 22 at PIT 34Sun, Oct 11, 1:00 PMPIT 59%✓ won
- CLE 12 at NYJ 17Sun, Oct 11, 1:00 PMNYJ 54%✓ won
- MIN 23 at NO 13Sun, Oct 11, 1:00 PMMIN 59%✓ won
- LV 17 at NE 45Sun, Oct 11, 1:00 PMNE 65%✓ won
- PHI 17 at JAX 19Sun, Oct 11, 9:30 AMJAX 74%✓ won
- TB 24 at DAL 16Thu, Oct 8, 8:15 PMDAL 79%✗ lost
- ATL 45 at NO 24Mon, Oct 5, 8:15 PMNO 52%✗ lost
- DET 26 at CAR 32Sun, Oct 4, 8:20 PMDET 65%✗ lost
- DEN 14 at SF 24Sun, Oct 4, 4:25 PMSF 60%✓ won
- LAC 23 at SEA 30Sun, Oct 4, 4:25 PMSEA 77%✓ won
- KC 30 at LV 27Sun, Oct 4, 4:25 PMKC 65%✓ won
- MIA 10 at MIN 15Sun, Oct 4, 4:05 PMMIN 81%✓ won
- ARI 24 at NYG 36Sun, Oct 4, 1:00 PMARI 53%✗ lost
- NE 29 at BUF 26Sun, Oct 4, 1:00 PMBUF 74%✗ lost
- NYJ 12 at CHI 23Sun, Oct 4, 1:00 PMCHI 65%✓ won
- JAX 22 at CIN 17Sun, Oct 4, 1:00 PMCIN 55%✗ lost
- LAR 24 at PHI 20Sun, Oct 4, 1:00 PMLAR 60%✓ won
- GB 17 at TB 14Sun, Oct 4, 1:00 PMGB 58%✓ won
- TEN 18 at BAL 24Sun, Oct 4, 1:00 PMBAL 85%✓ won
- DAL 34 at HOU 30Sun, Oct 4, 1:00 PMHOU 60%✗ lost
- IND 30 at WSH 13Sun, Oct 4, 9:30 AMIND 63%✓ won
- PIT 24 at CLE 27Thu, Oct 1, 8:15 PMPIT 57%✗ lost
- PHI 7 at CHI 27Mon, Sep 28, 8:15 PMPHI 63%✗ lost
By weekWon–lost week by week, locked picks only, with each week's 90% range.
Backtest winnersEvery game priced only from what was known before it, then scored once. Kept apart from the record.
Walk-forward backtest of the model alone, 2020–25 seasons: not in the recordWon 1,079. Its stated probabilities expected 1,100 (90% range 1,068–1,131): inside that range, in line with its probabilities.
By seasonEvery game priced only from what was known before it, reported in full.
What it says: in the walk-forward backtest the model called 63.7% of 1,693 NFL games, and the lanes above show whether its percentages hold up on the locked picks. The model makes no spread or total pick; its margin and total on each card are context beside the market’s numbers. The winner record, set beside the betting favorite’s, is the product.
Picks grade as each game goes final. For information only, not betting advice. 21+ only. Gambling problem? Call or text 1-800-GAMBLER.