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 · 2020
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
- 167
- 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 · 2020 · full board
| # | Player | Team | Pos | dropbacks | Raw | Shrunk | 90% interval |
|---|---|---|---|---|---|---|---|
| 1 | Josh Allen | BUF | QB | 599 | 55.1% | 53.3% | 50.4% – 56.3% |
| 2 | Aaron Rodgers | GB | QB | 543 | 54.0% | 52.3% | 49.3% – 55.4% |
| 3 | Patrick Mahomes | KC | QB | 612 | 53.1% | 51.8% | 48.9% – 54.7% |
| 4 | Ryan Fitzpatrick | MIA | QB | 282 | 53.2% | 50.9% | 47.0% – 54.8% |
| 5 | Dak Prescott | DAL | QB | 233 | 53.6% | 50.9% | 46.8% – 55.0% |
| 6 | Mitchell Trubisky | CHI | QB | 316 | 52.8% | 50.8% | 47.1% – 54.6% |
| 7 | Ryan Tannehill | TEN | QB | 507 | 51.9% | 50.7% | 47.5% – 53.8% |
| 8 | Deshaun Watson | HOU | QB | 595 | 51.6% | 50.6% | 47.6% – 53.6% |
| 9 | Kirk Cousins | MIN | QB | 559 | 51.3% | 50.3% | 47.3% – 53.4% |
| 10 | Baker Mayfield | CLE | QB | 508 | 50.6% | 49.7% | 46.6% – 52.9% |
| 11 | Drew Brees | NO | QB | 402 | 50.0% | 49.1% | 45.7% – 52.6% |
| 12 | Philip Rivers | IND | QB | 561 | 49.5% | 49.0% | 45.9% – 52.0% |
| 13 | Russell Wilson | SEA | QB | 606 | 49.5% | 49.0% | 46.0% – 51.9% |
| 14 | Tom Brady | TB | QB | 629 | 49.1% | 48.7% | 45.8% – 51.6% |
| 15 | Chad Henne◦ provisional | KC | QB | 40 | 55.0% | 48.6% | 42.9% – 54.3% |
| 16 | Derek Carr | LV | QB | 540 | 48.9% | 48.4% | 45.4% – 51.5% |
| 17 | Jared Goff | LA | QB | 574 | 48.3% | 48.0% | 45.0% – 51.0% |
| 18 | Marcus Mariota◦ provisional | LV | QB | 27 | 51.8% | 47.7% | 41.8% – 53.6% |
| 19 | Jimmy Garoppolo◦ provisional | SF | QB | 151 | 48.3% | 47.7% | 43.1% – 52.3% |
| 20 | Kyler Murray | ARI | QB | 579 | 47.8% | 47.7% | 44.7% – 50.7% |
| 21 | Matthew Stafford | DET | QB | 563 | 47.8% | 47.6% | 44.6% – 50.6% |
| 22 | Justin Herbert | LAC | QB | 626 | 47.4% | 47.4% | 44.5% – 50.3% |
| 23 | Joe Burrow | CIN | QB | 433 | 47.3% | 47.3% | 43.9% – 50.6% |
| 24 | Nick Mullens | SF | QB | 347 | 47.3% | 47.2% | 43.6% – 50.8% |
| 25 | Teddy Bridgewater | CAR | QB | 523 | 47.2% | 47.2% | 44.1% – 50.3% |
| 26 | Brandon Allen◦ provisional | CIN | QB | 149 | 47.0% | 47.0% | 42.4% – 51.6% |
| 27 | C.J. Beathard◦ provisional | SF | QB | 113 | 46.9% | 47.0% | 42.1% – 51.9% |
| 28 | Lamar Jackson | BAL | QB | 403 | 46.9% | 46.9% | 43.5% – 50.4% |
| 29 | Brett Rypien◦ provisional | DEN | QB | 41 | 46.3% | 46.9% | 41.2% – 52.6% |
| 30 | Ben Roethlisberger | PIT | QB | 625 | 46.6% | 46.7% | 43.8% – 49.6% |
| 31 | Matt Ryan | ATL | QB | 669 | 46.5% | 46.6% | 43.8% – 49.4% |
| 32 | Andy Dalton | DAL | QB | 357 | 46.2% | 46.5% | 42.9% – 50.1% |
| 33 | Cam Newton | NE | QB | 398 | 46.0% | 46.3% | 42.9% – 49.8% |
| 34 | Kyle Allen◦ provisional | WAS | QB | 96 | 44.8% | 46.2% | 41.2% – 51.3% |
| 35 | Tua Tagovailoa | MIA | QB | 310 | 45.5% | 46.0% | 42.3% – 49.8% |
| 36 | Tyrod Taylor◦ provisional | LAC | QB | 32 | 40.6% | 46.0% | 40.2% – 51.8% |
| 37 | Gardner Minshew | JAX | QB | 357 | 45.4% | 45.9% | 42.3% – 49.5% |
| 38 | Garrett Gilbert◦ provisional | DAL | QB | 39 | 41.0% | 45.9% | 40.2% – 51.6% |
| 39 | Brian Hoyer◦ provisional | NE | QB | 26 | 38.5% | 45.9% | 40.0% – 51.8% |
| 40 | Mike Glennon | JAX | QB | 189 | 43.9% | 45.4% | 41.1% – 49.7% |
| 41 | Chase Daniel◦ provisional | DET | QB | 45 | 37.8% | 45.1% | 39.5% – 50.7% |
| 42 | Mason Rudolph◦ provisional | PIT | QB | 45 | 37.8% | 45.1% | 39.5% – 50.7% |
| 43 | Ben DiNucci◦ provisional | DAL | QB | 49 | 36.7% | 44.7% | 39.2% – 50.3% |
| 44 | Jarrett Stidham◦ provisional | NE | QB | 48 | 35.4% | 44.5% | 38.9% – 50.0% |
| 45 | PJ Walker◦ provisional | CAR | QB | 60 | 36.7% | 44.3% | 38.9% – 49.7% |
| 46 | John Wolford◦ provisional | LA | QB | 40 | 32.5% | 44.2% | 38.6% – 49.9% |
| 47 | Colt McCoy◦ provisional | NYG | QB | 72 | 37.5% | 44.2% | 38.9% – 49.5% |
| 48 | Alex Smith | WAS | QB | 274 | 41.6% | 43.7% | 39.8% – 47.6% |
| 49 | Daniel Jones | NYG | QB | 494 | 42.3% | 43.5% | 40.4% – 46.7% |
| 50 | Ryan Finley◦ provisional | CIN | QB | 41 | 26.8% | 43.1% | 37.5% – 48.7% |
| 51 | Jeff Driskel◦ provisional | DEN | QB | 76 | 34.2% | 43.0% | 37.9% – 48.3% |
| 52 | Joe Flacco◦ provisional | NYJ | QB | 141 | 37.6% | 42.7% | 38.1% – 47.4% |
| 53 | Jake Luton◦ provisional | JAX | QB | 118 | 36.4% | 42.7% | 37.9% – 47.5% |
| 54 | Nick Foles | CHI | QB | 330 | 40.0% | 42.4% | 38.7% – 46.0% |
| 55 | Jalen Hurts◦ provisional | PHI | QB | 161 | 37.3% | 42.3% | 37.8% – 46.8% |
| 56 | Drew Lock | DEN | QB | 464 | 40.3% | 42.1% | 38.9% – 45.3% |
| 57 | Sam Darnold | NYJ | QB | 398 | 39.7% | 41.9% | 38.5% – 45.3% |
| 58 | Dwayne Haskins | WAS | QB | 262 | 37.4% | 41.2% | 37.3% – 45.1% |
| 59 | Carson Wentz | PHI | QB | 492 | 38.6% | 40.8% | 37.6% – 43.9% |
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).