The record
A projection is only worth what it would have told you before the game. Every number since week 3 is logged before kickoff, graded after the box score, and compared with the call you would make without us: each player's own average.
Replayed over weeks already played: clears the bar and misses by less than each player's own average. The live record is still accruing.
Replayed over 13,925 player-weeks that had already been played: correlation 0.83, average miss 2.2 opportunities against 2.4 for each player's own running average on the same player-weeks.
- At least 300 settled player-weeks per position, logged before kickoff
- Across at least 4 different weeks
- A correlation of 0.55 or better, per position
- Beating each player's own average on at least 150 paired rows
- "Tested", one step below: a replay of 2,000+ player-weeks at 0.60+ correlation.
The live ledger
Only rows logged before kickoff count. A week settles once, after every game in it is final.
- 2026 week 2Logged after its games: shown, never counted.746logged0before kickoff746settled0counted
- 2026 week 3Settled.810logged810before kickoff810settled387counted
- 2026 week 4Settled.889logged889before kickoff889settled390counted
- 2026 week 5Settled.834logged834before kickoff834settled385counted
- 2026 week 6Settled.798logged798before kickoff798settled353counted
- 2026 week 7Settled.791logged791before kickoff791settled353counted
- 2026 week 8Settled.760logged760before kickoff760settled354counted
- 2026 week 9Settled.833logged833before kickoff833settled380counted
- 2026 week 10Logged before kickoff; settles when the week is final.793logged793before kickoff0settled0counted
| Position | Settled | Weeks | Correlation | Avg miss | Bias | Our miss, paired | Baseline miss | Better by |
|---|---|---|---|---|---|---|---|---|
| QB | 242 | 7 | 0.55 | 1.63 | +0.06 | 1.70 | 1.85 | +0.15 |
| RB | 646 | 7 | 0.85 | 3.24 | −0.81 | 3.33 | 3.69 | +0.36 |
| TE | 682 | 7 | 0.76 | 1.30 | −0.06 | 1.31 | 1.39 | +0.07 |
| WR | 1,032 | 7 | 0.74 | 1.80 | −0.30 | 1.82 | 1.98 | +0.16 |
Only players who took the field count; a player who left hurt is out (the rule). Better by = the baseline's miss minus ours, on paired rows.
Still short: QB 242 of 300 rows, correlation 0.55, needs 0.55.
The replay
The engine the site runs, replayed over 2024, 2025 and 2026 so far, regular season only, on each week's real roster and injury report. Every number below is over the player-weeks where he played, the strip and the tables included; the ones he sat, and the ones where he left the game injured (the same rule as the live record above), are counted apart under the weekly strip. "Better by" is how many fewer opportunities a player-week it missed by than each player's own average.
Won 40 of 41 weeks against each player's own average.
beat the baseline lost to it. Left out of every number here: 3,437 player-weeks he did not play and 132 he left injured (our average miss on those, 3.8 opportunities). The share model was fitted on 2024-25, so those seasons are in-sample for it; 2026 is not. The replay has no spread or inactives list, and knows the starting quarterback after the fact, which flatters every version equally.
| By position | Corr | Avg miss | Better by |
|---|---|---|---|
| QB | 0.53 | 1.78 | +0.04 |
| RB | 0.82 | 3.42 | +0.47 |
| TE | 0.71 | 1.46 | +0.09 |
| WR | 0.72 | 1.85 | +0.14 |
| By sample | Corr | Avg miss | Better by |
|---|---|---|---|
| 1 game of usage | 0.81 | 2.03 | +0.25 |
| 2-3 of usage | 0.84 | 2.21 | +0.30 |
| 4-6 of usage | 0.83 | 2.19 | +0.25 |
| 7+ of usage | 0.83 | 2.20 | +0.16 |
The quarterback engine: proven
206 starts logged before kickoff over 7 weeks: passing yards missed by 58 on average, 9.5 less than each quarterback's own average.
What each piece earns
- 2024: average miss on passing yards
- Volume only61.7
- + the quarterback60.1
- + protection vs rush59.9
- + coverage (full model)60.2
- His own average, the call to beat67.4
- 2025: average miss on passing yards
- Volume only62.1
- + the quarterback60.2
- + protection vs rush60.1
- + coverage (full model)59.3
- His own average, the call to beat64.7
- 2026: average miss on passing yards
- Volume only61.4
- + the quarterback58.6
- + protection vs rush58.3
- + coverage (full model)58.3
- His own average, the call to beat65.5
Lower is better. Each row switches one more piece of the engine on.
How close it came
| Passing yards | Starts | Avg miss | Own avg | Better by |
|---|---|---|---|---|
| 2024 | 469 | 60.2 | 67.4 | +7.2 |
| 2025 | 466 | 59.3 | 64.7 | +5.4 |
| 2026 | 222 | 58.3 | 65.5 | +7.2 |
| The rest | Avg miss | Own avg | Better by |
|---|---|---|---|
| Attempts | 6.78 | 7.69 | +0.91 |
| Completions | 4.77 | 5.31 | +0.54 |
| Touchdown passes | 0.93 | 1.00 | +0.07 |
| Interceptions | 0.68 | 0.71 | +0.03 |
One quarterback's yards in one game swing on a few throws, so the test that matters is the miss against the call you would make without us.