NFL · Empirical Bayes
True-Talent Leaderboards
A raw single-season rate is a noisy guess at a player’s real ability, and the noise is worst for the smallest samples — so a naïve leaderboard is topped by whoever got lucky in the fewest tries. This one shrinks every rate toward its position-group prior by how much the sample can be trusted, and shows a 90% credible interval instead of a point. Toggle raw vs. shrunk to watch the flukes fall back to the pack.
EPA per dropback · 2025
Mean QB EPA per dropback.
- Beats raw by
- 14.9%
- lower out-of-sample error
- RMSE raw → shrunk
- 0.17 → 0.14
- odd vs. even weeks
- Split-half reliability
- 0.47
- how repeatable the raw stat is
- Stabilizes at
- 196
- dropbacks to trust the number
Each row is a player: the solid dot is the shrunk estimate, the hollow dot the raw rate, joined by the red pull of regression; the grey bar is the 90% credible interval. A hollow shrunk dot means the sample is below the stabilization line — the number is mostly the position prior. Switch to Raw rank and watch the small-sample names climb.
EPA per dropback · 2025 · full board
| # | Player | Team | Pos | dropbacks | Raw | Shrunk | 90% interval |
|---|---|---|---|---|---|---|---|
| 1 | Drake Maye | NE | QB | 540 | +0.31 | +0.24 | +0.14 – +0.33 |
| 2 | Matthew Stafford | LA | QB | 617 | +0.24 | +0.20 | +0.11 – +0.29 |
| 3 | Jordan Love | GB | QB | 461 | +0.24 | +0.18 | +0.08 – +0.28 |
| 4 | Malik Willis◦ provisional | GB | QB | 38 | +0.74 | +0.16 | -0.01 – +0.33 |
| 5 | Mitchell Trubisky◦ provisional | BUF | QB | 37 | +0.70 | +0.15 | -0.02 – +0.32 |
| 6 | Dak Prescott | DAL | QB | 631 | +0.18 | +0.15 | +0.06 – +0.24 |
| 7 | Brock Purdy | SF | QB | 298 | +0.21 | +0.15 | +0.03 – +0.26 |
| 8 | Jared Goff | DET | QB | 614 | +0.17 | +0.14 | +0.05 – +0.23 |
| 9 | Daniel Jones | IND | QB | 409 | +0.16 | +0.12 | +0.02 – +0.23 |
| 10 | Sam Darnold | SEA | QB | 502 | +0.15 | +0.12 | +0.02 – +0.22 |
| 11 | Josh Allen | BUF | QB | 506 | +0.14 | +0.11 | +0.02 – +0.21 |
| 12 | Mac Jones | SF | QB | 304 | +0.15 | +0.11 | -0.01 – +0.22 |
| 13 | Patrick Mahomes | KC | QB | 537 | +0.13 | +0.11 | +0.01 – +0.20 |
| 14 | Joe Burrow | CIN | QB | 278 | +0.13 | +0.10 | -0.02 – +0.21 |
| 15 | Jameis Winston◦ provisional | NYG | QB | 69 | +0.22 | +0.09 | -0.07 – +0.25 |
| 16 | C.J. Stroud | HOU | QB | 446 | +0.11 | +0.09 | -0.01 – +0.19 |
| 17 | Bo Nix | DEN | QB | 638 | +0.10 | +0.09 | -0.00 – +0.18 |
| 18 | Caleb Williams | CHI | QB | 594 | +0.07 | +0.06 | -0.03 – +0.16 |
| 19 | Jalen Hurts | PHI | QB | 486 | +0.07 | +0.06 | -0.04 – +0.16 |
| 20 | Trevor Lawrence | JAX | QB | 600 | +0.06 | +0.06 | -0.03 – +0.15 |
| 21 | Tyler Huntley◦ provisional | BAL | QB | 74 | +0.08 | +0.05 | -0.10 – +0.21 |
| 22 | Josh Johnson◦ provisional | WAS | QB | 54 | +0.08 | +0.05 | -0.11 – +0.22 |
| 23 | Lamar Jackson | BAL | QB | 338 | +0.05 | +0.05 | -0.06 – +0.16 |
| 24 | Aaron Rodgers | PIT | QB | 527 | +0.05 | +0.05 | -0.05 – +0.15 |
| 25 | Mason Rudolph◦ provisional | PIT | QB | 54 | +0.05 | +0.05 | -0.12 – +0.21 |
| 26 | Carson Wentz◦ provisional | MIN | QB | 188 | +0.03 | +0.04 | -0.09 – +0.17 |
| 27 | Kyler Murray◦ provisional | ARI | QB | 177 | +0.03 | +0.04 | -0.10 – +0.17 |
| 28 | Tua Tagovailoa | MIA | QB | 415 | +0.03 | +0.04 | -0.07 – +0.14 |
| 29 | Justin Herbert | LAC | QB | 566 | +0.03 | +0.03 | -0.06 – +0.13 |
| 30 | Michael Penix Jr. | ATL | QB | 290 | +0.02 | +0.03 | -0.09 – +0.15 |
| 31 | Tanner McKee◦ provisional | PHI | QB | 46 | -0.05 | +0.03 | -0.14 – +0.19 |
| 32 | Quinn Ewers◦ provisional | MIA | QB | 91 | -0.02 | +0.02 | -0.13 – +0.18 |
| 33 | Jaxson Dart | NYG | QB | 375 | +0.01 | +0.02 | -0.09 – +0.13 |
| 34 | Tyler Shough | NO | QB | 356 | +0.01 | +0.02 | -0.09 – +0.13 |
| 35 | Marcus Mariota | WAS | QB | 244 | -0.01 | +0.02 | -0.11 – +0.14 |
| 36 | Kirk Cousins | ATL | QB | 281 | -0.01 | +0.01 | -0.10 – +0.13 |
| 37 | Andy Dalton◦ provisional | CAR | QB | 45 | -0.13 | +0.01 | -0.15 – +0.18 |
| 38 | Baker Mayfield | TB | QB | 584 | +0.00 | +0.01 | -0.08 – +0.10 |
| 39 | Jayden Daniels | WAS | QB | 205 | -0.02 | +0.01 | -0.12 – +0.14 |
| 40 | Jacoby Brissett | ARI | QB | 530 | -0.01 | +0.01 | -0.09 – +0.10 |
| 41 | Riley Leonard◦ provisional | IND | QB | 70 | -0.12 | +0.00 | -0.16 – +0.16 |
| 42 | Philip Rivers◦ provisional | IND | QB | 97 | -0.12 | -0.01 | -0.16 – +0.14 |
| 43 | Tyrod Taylor◦ provisional | NYJ | QB | 147 | -0.09 | -0.01 | -0.15 – +0.13 |
| 44 | Bryce Young | CAR | QB | 506 | -0.04 | -0.02 | -0.12 – +0.08 |
| 45 | Russell Wilson◦ provisional | NYG | QB | 129 | -0.12 | -0.02 | -0.16 – +0.12 |
| 46 | Davis Mills◦ provisional | HOU | QB | 169 | -0.10 | -0.02 | -0.16 – +0.11 |
| 47 | Spencer Rattler | NO | QB | 272 | -0.08 | -0.03 | -0.15 – +0.09 |
| 48 | Justin Fields | NYJ | QB | 233 | -0.09 | -0.03 | -0.16 – +0.09 |
| 49 | Cooper Rush◦ provisional | BAL | QB | 53 | -0.36 | -0.04 | -0.20 – +0.12 |
| 50 | Joe Flacco | CIN | QB | 432 | -0.10 | -0.05 | -0.15 – +0.05 |
| 51 | Brandon Allen◦ provisional | TEN | QB | 31 | -0.78 | -0.07 | -0.24 – +0.10 |
| 52 | Kenny Pickett◦ provisional | LV | QB | 52 | -0.55 | -0.08 | -0.24 – +0.08 |
| 53 | Chris Oladokun◦ provisional | KC | QB | 62 | -0.48 | -0.08 | -0.24 – +0.08 |
| 54 | Jake Browning◦ provisional | CIN | QB | 134 | -0.29 | -0.09 | -0.23 – +0.05 |
| 55 | Dillon Gabriel | CLE | QB | 204 | -0.22 | -0.09 | -0.22 – +0.04 |
| 56 | Geno Smith | LV | QB | 503 | -0.15 | -0.09 | -0.19 – +0.01 |
| 57 | J.J. McCarthy | MIN | QB | 273 | -0.20 | -0.10 | -0.22 – +0.02 |
| 58 | Trey Lance◦ provisional | LAC | QB | 59 | -0.65 | -0.12 | -0.28 – +0.05 |
| 59 | Shedeur Sanders | CLE | QB | 235 | -0.25 | -0.12 | -0.24 – +0.01 |
| 60 | Cam Ward | TEN | QB | 592 | -0.18 | -0.13 | -0.22 – -0.04 |
| 61 | Brady Cook◦ provisional | NYJ | QB | 173 | -0.40 | -0.16 | -0.30 – -0.03 |
| 62 | Max Brosmer◦ provisional | MIN | QB | 85 | -0.67 | -0.17 | -0.32 – -0.02 |
How the shrinkage works
Two estimators
Rate stats (completion %, success rate, catch rate) use a beta-binomial model: a Beta(α, β) prior fit by marginal likelihood over each position group, then a Beta posterior per player. Per-play averages (EPA, CPOE, yards) use a normal-normal model with DerSimonian–Laird between-player variance. Both pull each player toward their group by exactly how thin their sample is.
Does it help? & the fine print
The trust panel’s numbers come from a leakage-free odd/even-week holdout: fit on odd weeks, predict even-week raw. Shrinkage lowers out-of-sample error for every stat. Priors are fit per season and per position (WR and TE separately), so a TE’s baseline isn’t a skill. Regular season only, 2016–2025. Rushing and receiving efficiency are heavily scheme-driven — read the wide bands as the honesty they are.
Source: nflverse play-by-play. Counted and computed deterministically — never modeled by a language model. Built by build_nfl_leaderboards.py (byte-reproducible).