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 · 2017
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
- 186
- 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 · 2017 · full board
| # | Player | Team | Pos | dropbacks | Raw | Shrunk | 90% interval |
|---|---|---|---|---|---|---|---|
| 1 | Tom Brady | NE | QB | 618 | 50.5% | 48.8% | 45.9% – 51.7% |
| 2 | Jimmy Garoppolo◦ provisional | SF | QB | 184 | 53.8% | 48.5% | 44.2% – 52.7% |
| 3 | Jameis Winston | TB | QB | 474 | 50.2% | 48.2% | 45.0% – 51.4% |
| 4 | Ben Roethlisberger | PIT | QB | 582 | 49.7% | 48.1% | 45.1% – 51.0% |
| 5 | Drew Brees | NO | QB | 553 | 49.5% | 47.9% | 44.9% – 51.0% |
| 6 | Matt Ryan | ATL | QB | 550 | 48.7% | 47.3% | 44.3% – 50.3% |
| 7 | Philip Rivers | LAC | QB | 592 | 47.8% | 46.7% | 43.8% – 49.6% |
| 8 | Alex Smith | KC | QB | 540 | 47.4% | 46.3% | 43.3% – 49.4% |
| 9 | Carson Wentz | PHI | QB | 473 | 47.4% | 46.2% | 43.0% – 49.4% |
| 10 | Aaron Rodgers | GB | QB | 261 | 47.5% | 45.7% | 41.8% – 49.6% |
| 11 | Case Keenum | MIN | QB | 504 | 46.6% | 45.7% | 42.6% – 48.8% |
| 12 | Deshaun Watson | HOU | QB | 224 | 47.3% | 45.4% | 41.4% – 49.5% |
| 13 | Patrick Mahomes◦ provisional | KC | QB | 37 | 56.8% | 45.4% | 40.0% – 50.9% |
| 14 | Blake Bortles | JAX | QB | 550 | 45.3% | 44.7% | 41.7% – 47.8% |
| 15 | Matthew Stafford | DET | QB | 615 | 45.0% | 44.6% | 41.7% – 47.5% |
| 16 | Jared Goff | LA | QB | 500 | 44.8% | 44.4% | 41.3% – 47.5% |
| 17 | Russell Wilson | SEA | QB | 597 | 44.6% | 44.2% | 41.3% – 47.1% |
| 18 | Ryan Fitzpatrick◦ provisional | TB | QB | 172 | 45.4% | 44.2% | 39.9% – 48.5% |
| 19 | Josh McCown | NYJ | QB | 432 | 44.2% | 43.9% | 40.6% – 47.2% |
| 20 | Kirk Cousins | WAS | QB | 581 | 43.9% | 43.7% | 40.8% – 46.7% |
| 21 | David Fales◦ provisional | MIA | QB | 45 | 44.4% | 43.4% | 38.1% – 48.8% |
| 22 | Carson Palmer | ARI | QB | 290 | 43.5% | 43.3% | 39.6% – 47.1% |
| 23 | Derek Carr | LV | QB | 537 | 43.4% | 43.3% | 40.3% – 46.4% |
| 24 | Eli Manning | NYG | QB | 607 | 43.3% | 43.3% | 40.4% – 46.2% |
| 25 | Marcus Mariota | TEN | QB | 479 | 43.0% | 43.0% | 39.9% – 46.2% |
| 26 | Landry Jones◦ provisional | PIT | QB | 31 | 41.9% | 43.0% | 37.5% – 48.5% |
| 27 | Sam Bradford◦ provisional | MIN | QB | 48 | 41.7% | 42.9% | 37.6% – 48.2% |
| 28 | Paxton Lynch◦ provisional | DEN | QB | 54 | 40.7% | 42.6% | 37.4% – 47.9% |
| 29 | EJ Manuel◦ provisional | LV | QB | 47 | 40.4% | 42.6% | 37.3% – 48.0% |
| 30 | Cam Newton | CAR | QB | 527 | 42.3% | 42.5% | 39.5% – 45.6% |
| 31 | Kevin Hogan◦ provisional | CLE | QB | 81 | 40.7% | 42.4% | 37.5% – 47.4% |
| 32 | Nathan Peterman◦ provisional | BUF | QB | 48 | 39.6% | 42.4% | 37.2% – 47.8% |
| 33 | Geno Smith◦ provisional | NYG | QB | 39 | 38.5% | 42.4% | 37.0% – 47.8% |
| 34 | Dak Prescott | DAL | QB | 523 | 41.9% | 42.2% | 39.2% – 45.3% |
| 35 | Nate Sudfeld◦ provisional | PHI | QB | 26 | 34.6% | 42.1% | 36.6% – 47.7% |
| 36 | Jay Cutler | MIA | QB | 449 | 41.6% | 42.1% | 38.9% – 45.3% |
| 37 | Matt Moore◦ provisional | MIA | QB | 138 | 40.6% | 42.1% | 37.6% – 46.6% |
| 38 | Joe Flacco | BAL | QB | 574 | 41.6% | 42.0% | 39.1% – 45.0% |
| 39 | Mike Glennon◦ provisional | CHI | QB | 148 | 39.9% | 41.7% | 37.3% – 46.2% |
| 40 | Tyrod Taylor | BUF | QB | 466 | 40.6% | 41.3% | 38.1% – 44.5% |
| 41 | Andy Dalton | CIN | QB | 534 | 40.6% | 41.3% | 38.3% – 44.3% |
| 42 | Nick Foles◦ provisional | PHI | QB | 106 | 37.7% | 41.2% | 36.5% – 46.0% |
| 43 | Cody Kessler◦ provisional | CLE | QB | 29 | 27.6% | 41.1% | 35.6% – 46.6% |
| 44 | Tom Savage | HOU | QB | 243 | 39.1% | 40.9% | 37.0% – 44.8% |
| 45 | Jacoby Brissett | IND | QB | 519 | 39.9% | 40.8% | 37.7% – 43.8% |
| 46 | Trevor Siemian | DEN | QB | 384 | 39.3% | 40.6% | 37.2% – 44.0% |
| 47 | Brock Osweiler◦ provisional | DEN | QB | 182 | 37.9% | 40.6% | 36.4% – 44.8% |
| 48 | Brian Hoyer | NE | QB | 228 | 38.2% | 40.4% | 36.5% – 44.4% |
| 49 | Matt Cassel◦ provisional | TEN | QB | 50 | 30.0% | 40.4% | 35.2% – 45.7% |
| 50 | Brett Hundley | GB | QB | 346 | 38.7% | 40.3% | 36.8% – 43.8% |
| 51 | Mitchell Trubisky | CHI | QB | 361 | 38.8% | 40.3% | 36.9% – 43.7% |
| 52 | Drew Stanton◦ provisional | ARI | QB | 166 | 36.8% | 40.1% | 35.9% – 44.5% |
| 53 | Sean Mannion◦ provisional | LA | QB | 40 | 25.0% | 40.0% | 34.7% – 45.4% |
| 54 | Blaine Gabbert | ARI | QB | 194 | 36.6% | 39.8% | 35.7% – 44.0% |
| 55 | C.J. Beathard | SF | QB | 242 | 37.2% | 39.8% | 35.9% – 43.7% |
| 56 | Bryce Petty◦ provisional | NYJ | QB | 121 | 33.9% | 39.5% | 35.0% – 44.1% |
| 57 | T.J. Yates◦ provisional | HOU | QB | 111 | 32.4% | 39.2% | 34.5% – 43.9% |
| 58 | DeShone Kizer | CLE | QB | 513 | 33.9% | 36.4% | 33.4% – 39.4% |
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).