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.
Dropback success rate · 2025
Share of dropbacks (incl. sacks & scrambles) with positive EPA.
- Beats raw by
- 15.6%
- lower out-of-sample error
- RMSE raw → shrunk
- 5.5% → 4.6%
- odd vs. even weeks
- Split-half reliability
- 0.47
- how repeatable the raw stat is
- Stabilizes at
- 143
- 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.
Dropback success rate · 2025 · full board
| # | Player | Team | Pos | dropbacks | Raw | Shrunk | 90% interval |
|---|---|---|---|---|---|---|---|
| 1 | Drake Maye | NE | QB | 540 | 54.4% | 52.4% | 49.3% – 55.5% |
| 2 | Matthew Stafford | LA | QB | 617 | 53.2% | 51.6% | 48.6% – 54.5% |
| 3 | Brock Purdy | SF | QB | 298 | 53.4% | 50.5% | 46.6% – 54.4% |
| 4 | Sam Darnold | SEA | QB | 502 | 51.8% | 50.2% | 47.0% – 53.4% |
| 5 | Mac Jones | SF | QB | 304 | 52.3% | 49.9% | 46.0% – 53.7% |
| 6 | Malik Willis◦ provisional | GB | QB | 38 | 65.8% | 49.1% | 43.0% – 55.2% |
| 7 | Jordan Love | GB | QB | 461 | 49.7% | 48.5% | 45.1% – 51.8% |
| 8 | Daniel Jones | IND | QB | 409 | 49.6% | 48.3% | 44.9% – 51.8% |
| 9 | Mitchell Trubisky◦ provisional | BUF | QB | 37 | 62.2% | 48.2% | 42.1% – 54.4% |
| 10 | Jared Goff | DET | QB | 614 | 48.4% | 47.7% | 44.7% – 50.6% |
| 11 | Joe Burrow | CIN | QB | 278 | 48.9% | 47.5% | 43.5% – 51.5% |
| 12 | Dak Prescott | DAL | QB | 631 | 47.9% | 47.3% | 44.3% – 50.2% |
| 13 | Patrick Mahomes | KC | QB | 537 | 47.9% | 47.2% | 44.0% – 50.3% |
| 14 | Carson Wentz | MIN | QB | 188 | 48.9% | 47.1% | 42.6% – 51.6% |
| 15 | Marcus Mariota | WAS | QB | 244 | 47.1% | 46.2% | 42.1% – 50.4% |
| 16 | Josh Allen | BUF | QB | 506 | 46.6% | 46.2% | 43.0% – 49.4% |
| 17 | Josh Johnson◦ provisional | WAS | QB | 54 | 50.0% | 46.1% | 40.3% – 52.0% |
| 18 | Mason Rudolph◦ provisional | PIT | QB | 54 | 50.0% | 46.1% | 40.3% – 52.0% |
| 19 | Trevor Lawrence | JAX | QB | 600 | 46.2% | 45.9% | 42.9% – 48.9% |
| 20 | Tyler Huntley◦ provisional | BAL | QB | 74 | 47.3% | 45.5% | 40.0% – 51.1% |
| 21 | Tanner McKee◦ provisional | PHI | QB | 46 | 47.8% | 45.4% | 39.5% – 51.4% |
| 22 | C.J. Stroud | HOU | QB | 446 | 45.5% | 45.3% | 41.9% – 48.7% |
| 23 | Lamar Jackson | BAL | QB | 338 | 45.6% | 45.3% | 41.6% – 49.0% |
| 24 | Justin Herbert | LAC | QB | 566 | 45.2% | 45.1% | 42.0% – 48.2% |
| 25 | Michael Penix Jr. | ATL | QB | 290 | 45.2% | 45.0% | 41.1% – 48.9% |
| 26 | Tua Tagovailoa | MIA | QB | 415 | 44.8% | 44.8% | 41.3% – 48.2% |
| 27 | Jacoby Brissett | ARI | QB | 530 | 44.7% | 44.7% | 41.6% – 47.9% |
| 28 | Andy Dalton◦ provisional | CAR | QB | 45 | 44.4% | 44.6% | 38.7% – 50.6% |
| 29 | Jayden Daniels | WAS | QB | 205 | 44.4% | 44.5% | 40.1% – 48.9% |
| 30 | Kirk Cousins | ATL | QB | 281 | 44.1% | 44.3% | 40.4% – 48.3% |
| 31 | Spencer Rattler | NO | QB | 272 | 44.1% | 44.3% | 40.3% – 48.3% |
| 32 | Bo Nix | DEN | QB | 638 | 44.2% | 44.3% | 41.4% – 47.2% |
| 33 | Jameis Winston◦ provisional | NYG | QB | 69 | 43.5% | 44.3% | 38.7% – 49.9% |
| 34 | Tyler Shough | NO | QB | 356 | 43.8% | 44.0% | 40.4% – 47.7% |
| 35 | Kyler Murray | ARI | QB | 177 | 43.5% | 44.0% | 39.5% – 48.6% |
| 36 | Cooper Rush◦ provisional | BAL | QB | 53 | 41.5% | 43.8% | 38.0% – 49.6% |
| 37 | Tyrod Taylor | NYJ | QB | 147 | 42.9% | 43.7% | 39.0% – 48.5% |
| 38 | Philip Rivers◦ provisional | IND | QB | 97 | 42.3% | 43.7% | 38.5% – 49.0% |
| 39 | Jalen Hurts | PHI | QB | 486 | 43.2% | 43.5% | 40.3% – 46.8% |
| 40 | Caleb Williams | CHI | QB | 594 | 43.3% | 43.5% | 40.5% – 46.5% |
| 41 | Baker Mayfield | TB | QB | 584 | 43.1% | 43.4% | 40.4% – 46.5% |
| 42 | Aaron Rodgers | PIT | QB | 527 | 43.1% | 43.4% | 40.3% – 46.6% |
| 43 | Bryce Young | CAR | QB | 506 | 42.7% | 43.1% | 39.9% – 46.3% |
| 44 | Quinn Ewers◦ provisional | MIA | QB | 91 | 40.7% | 43.1% | 37.8% – 48.4% |
| 45 | Geno Smith | LV | QB | 503 | 42.1% | 42.7% | 39.5% – 45.9% |
| 46 | Justin Fields | NYJ | QB | 233 | 41.2% | 42.5% | 38.3% – 46.7% |
| 47 | Jaxson Dart | NYG | QB | 375 | 41.6% | 42.4% | 38.9% – 46.0% |
| 48 | Davis Mills | HOU | QB | 169 | 39.6% | 41.9% | 37.4% – 46.6% |
| 49 | Joe Flacco | CIN | QB | 432 | 41.0% | 41.9% | 38.5% – 45.3% |
| 50 | J.J. McCarthy | MIN | QB | 273 | 40.3% | 41.8% | 37.8% – 45.8% |
| 51 | Jake Browning◦ provisional | CIN | QB | 134 | 38.1% | 41.4% | 36.6% – 46.4% |
| 52 | Russell Wilson◦ provisional | NYG | QB | 129 | 36.4% | 40.7% | 35.9% – 45.7% |
| 53 | Riley Leonard◦ provisional | IND | QB | 70 | 31.4% | 40.3% | 34.8% – 45.9% |
| 54 | Kenny Pickett◦ provisional | LV | QB | 52 | 26.9% | 39.9% | 34.2% – 45.7% |
| 55 | Brandon Allen◦ provisional | TEN | QB | 31 | 16.1% | 39.6% | 33.5% – 45.7% |
| 56 | Trey Lance◦ provisional | LAC | QB | 59 | 27.1% | 39.5% | 33.9% – 45.2% |
| 57 | Chris Oladokun◦ provisional | KC | QB | 62 | 25.8% | 38.9% | 33.4% – 44.6% |
| 58 | Cam Ward | TEN | QB | 592 | 36.8% | 38.3% | 35.4% – 41.3% |
| 59 | Shedeur Sanders | CLE | QB | 235 | 33.6% | 37.8% | 33.7% – 41.9% |
| 60 | Max Brosmer◦ provisional | MIN | QB | 85 | 25.9% | 37.6% | 32.4% – 43.0% |
| 61 | Brady Cook | NYJ | QB | 173 | 31.2% | 37.3% | 32.9% – 41.8% |
| 62 | Dillon Gabriel | CLE | QB | 204 | 31.9% | 37.1% | 32.9% – 41.4% |
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).