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 · 2019
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
- 189
- 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 · 2019 · full board
| # | Player | Team | Pos | dropbacks | Raw | Shrunk | 90% interval |
|---|---|---|---|---|---|---|---|
| 1 | Drew Brees | NO | QB | 386 | 53.9% | 50.9% | 47.5% – 54.3% |
| 2 | Dak Prescott | DAL | QB | 620 | 52.3% | 50.5% | 47.6% – 53.4% |
| 3 | Ryan Tannehill | TEN | QB | 315 | 53.6% | 50.3% | 46.7% – 54.0% |
| 4 | Philip Rivers | LAC | QB | 623 | 50.2% | 49.0% | 46.1% – 51.9% |
| 5 | Jimmy Garoppolo | SF | QB | 514 | 50.0% | 48.6% | 45.5% – 51.7% |
| 6 | Matt Schaub◦ provisional | ATL | QB | 70 | 58.6% | 48.5% | 43.4% – 53.6% |
| 7 | Patrick Mahomes | KC | QB | 502 | 49.8% | 48.4% | 45.3% – 51.6% |
| 8 | Derek Carr | LV | QB | 541 | 49.4% | 48.2% | 45.1% – 51.2% |
| 9 | Russell Wilson | SEA | QB | 563 | 48.5% | 47.6% | 44.6% – 50.6% |
| 10 | Lamar Jackson | BAL | QB | 423 | 48.7% | 47.5% | 44.2% – 50.8% |
| 11 | Deshaun Watson | HOU | QB | 538 | 48.1% | 47.3% | 44.2% – 50.3% |
| 12 | Jameis Winston | TB | QB | 674 | 47.9% | 47.2% | 44.5% – 50.0% |
| 13 | Matthew Stafford | DET | QB | 308 | 48.7% | 47.2% | 43.5% – 50.9% |
| 14 | Kyle Allen | CAR | QB | 536 | 46.6% | 46.2% | 43.1% – 49.2% |
| 15 | Jared Goff | LA | QB | 649 | 46.5% | 46.1% | 43.3% – 49.0% |
| 16 | Kirk Cousins | MIN | QB | 475 | 46.3% | 45.9% | 42.7% – 49.1% |
| 17 | Carson Wentz | PHI | QB | 646 | 46.1% | 45.8% | 43.0% – 48.7% |
| 18 | Matt Ryan | ATL | QB | 668 | 46.1% | 45.8% | 43.0% – 48.6% |
| 19 | Chase Daniel◦ provisional | CHI | QB | 71 | 47.9% | 45.6% | 40.6% – 50.7% |
| 20 | Case Keenum | WAS | QB | 262 | 46.2% | 45.6% | 41.8% – 49.5% |
| 21 | Robert Griffin III◦ provisional | BAL | QB | 43 | 48.8% | 45.6% | 40.2% – 50.9% |
| 22 | Teddy Bridgewater | NO | QB | 208 | 46.2% | 45.5% | 41.4% – 49.6% |
| 23 | Ryan Fitzpatrick | MIA | QB | 543 | 45.5% | 45.3% | 42.3% – 48.3% |
| 24 | Matt Moore◦ provisional | KC | QB | 98 | 45.9% | 45.2% | 40.4% – 50.0% |
| 25 | Drew Lock◦ provisional | DEN | QB | 161 | 43.5% | 44.2% | 39.9% – 48.6% |
| 26 | AJ McCarron◦ provisional | HOU | QB | 42 | 40.5% | 44.0% | 38.7% – 49.4% |
| 27 | Jacoby Brissett | IND | QB | 472 | 43.6% | 44.0% | 40.8% – 47.2% |
| 28 | Eli Manning◦ provisional | NYG | QB | 152 | 42.8% | 43.9% | 39.5% – 48.3% |
| 29 | Aaron Rodgers | GB | QB | 606 | 43.6% | 43.9% | 41.0% – 46.8% |
| 30 | Matt Barkley◦ provisional | BUF | QB | 52 | 40.4% | 43.9% | 38.6% – 49.1% |
| 31 | Mitchell Trubisky | CHI | QB | 552 | 43.1% | 43.5% | 40.6% – 46.6% |
| 32 | Ben Roethlisberger◦ provisional | PIT | QB | 64 | 39.1% | 43.4% | 38.3% – 48.5% |
| 33 | Cam Newton◦ provisional | CAR | QB | 94 | 40.4% | 43.4% | 38.5% – 48.2% |
| 34 | Brian Hoyer◦ provisional | IND | QB | 71 | 39.4% | 43.3% | 38.3% – 48.4% |
| 35 | Joe Flacco | DEN | QB | 288 | 42.4% | 43.3% | 39.6% – 47.1% |
| 36 | Kyler Murray | ARI | QB | 589 | 42.8% | 43.3% | 40.4% – 46.2% |
| 37 | Sam Darnold | NYJ | QB | 475 | 42.5% | 43.2% | 40.0% – 46.4% |
| 38 | Andy Dalton | CIN | QB | 560 | 42.1% | 42.8% | 39.9% – 45.8% |
| 39 | Gardner Minshew | JAX | QB | 507 | 42.0% | 42.8% | 39.7% – 45.9% |
| 40 | Baker Mayfield | CLE | QB | 575 | 41.9% | 42.6% | 39.7% – 45.6% |
| 41 | Tom Brady | NE | QB | 642 | 41.9% | 42.6% | 39.8% – 45.4% |
| 42 | Josh Allen | BUF | QB | 498 | 41.6% | 42.5% | 39.4% – 45.6% |
| 43 | Jeff Driskel◦ provisional | DET | QB | 115 | 38.3% | 42.3% | 37.7% – 47.0% |
| 44 | Devlin Hodges◦ provisional | PIT | QB | 175 | 39.4% | 42.2% | 38.0% – 46.5% |
| 45 | Dwayne Haskins | WAS | QB | 234 | 39.3% | 41.8% | 37.9% – 45.7% |
| 46 | Will Grier◦ provisional | CAR | QB | 58 | 31.0% | 41.6% | 36.5% – 46.8% |
| 47 | Mason Rudolph | PIT | QB | 299 | 39.5% | 41.5% | 37.9% – 45.2% |
| 48 | Daniel Jones | NYG | QB | 500 | 40.2% | 41.5% | 38.4% – 44.6% |
| 49 | Colt McCoy◦ provisional | WAS | QB | 33 | 21.2% | 41.3% | 35.9% – 46.8% |
| 50 | Marcus Mariota◦ provisional | TEN | QB | 183 | 37.2% | 41.0% | 36.9% – 45.3% |
| 51 | Nick Foles◦ provisional | JAX | QB | 128 | 35.2% | 40.9% | 36.4% – 45.5% |
| 52 | David Blough◦ provisional | DET | QB | 188 | 34.6% | 39.7% | 35.6% – 43.9% |
| 53 | Brandon Allen◦ provisional | DEN | QB | 93 | 29.0% | 39.6% | 34.9% – 44.4% |
| 54 | Ryan Finley◦ provisional | CIN | QB | 98 | 29.6% | 39.6% | 34.9% – 44.4% |
| 55 | Josh Rosen◦ provisional | MIA | QB | 125 | 31.2% | 39.4% | 34.9% – 44.0% |
| 56 | Luke Falk◦ provisional | NYJ | QB | 90 | 25.6% | 38.6% | 33.9% – 43.4% |
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).