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.

Completion % · 2024

Completions per pass attempt. Depth-of-target confound: a checkdown offense completes more. Pair with CPOE.

Beats raw by
15.6%
lower out-of-sample error
RMSE raw → shrunk
5.2% → 4.4%
odd vs. even weeks
Split-half reliability
0.42
how repeatable the raw stat is
Stabilizes at
173
attempts to trust the number
shrunkrawthe shrink90% CI
50%55%60%65%70%avg1Jared GoffDET · 57166.4%2Tua TagovailoaMIA · 42366.3%3Baker MayfieldTB · 61365.0%4Joe BurrowCIN · 70264.4%5Kyler MurrayARI · 57164.0%6Geno SmithSEA · 63163.5%7Derek CarrNO · 29063.2%8Bo NixDEN · 59262.7%9Patrick MahomesKC · 61962.6%10Lamar JacksonBAL · 49962.5%11Marcus MariotaWAS · 4862.4%12Joe Milton IIINE · 2962.3%13Kirk CousinsATL · 48262.1%14Jayden DanielsWAS · 53061.9%15Mitchell TrubiskyBUF · 2661.7%16Matthew StaffordLA · 54761.7%17Andy DaltonCAR · 16861.6%18Jalen HurtsPHI · 40061.4%19Josh AllenBUF · 49861.2%20Joshua DobbsSF · 4961.2%21Brock PurdySF · 48761.2%22Tyrod TaylorNYJ · 2561.1%23Malik WillisGB · 6361.0%24Tanner McKeePHI · 4760.9%25Mac JonesJAX · 27960.8%26Sam DarnoldMIN · 59460.6%27Mason RudolphTEN · 24060.5%28Tommy DeVitoNYG · 5060.5%29Jordan LoveGB · 44260.5%30Justin HerbertLAC · 54860.5%31Joe FlaccoIND · 26760.4%

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.

Completion % · 2024 · full board

Completion % leaderboard for the 2024 NFL season, shrunk by empirical Bayes with 90% credible intervals, ranked by the shrunk estimate.
#PlayerTeamPosattemptsRawShrunk90% interval
1Jared GoffDETQB57168.3%66.4%63.5%69.2%
2Tua TagovailoaMIAQB42368.8%66.3%63.0%69.4%
3Baker MayfieldTBQB61366.4%65.0%62.2%67.8%
4Joe BurrowCINQB70265.5%64.4%61.8%67.1%
5Kyler MurrayARIQB57165.1%64.0%61.1%66.8%
6Geno SmithSEAQB63164.5%63.5%60.7%66.3%
7Derek CarrNOQB29065.2%63.2%59.5%66.9%
8Bo NixDENQB59263.5%62.7%59.8%65.6%
9Patrick MahomesKCQB61963.3%62.6%59.8%65.4%
10Lamar JacksonBALQB49963.3%62.5%59.4%65.5%
11Marcus Mariota◦ provisionalWASQB4870.8%62.4%57.0%67.7%
12Joe Milton III◦ provisionalNEQB2975.9%62.3%56.6%67.8%
13Kirk CousinsATLQB48262.9%62.1%59.0%65.2%
14Jayden DanielsWASQB53062.5%61.9%58.8%64.8%
15Mitchell Trubisky◦ provisionalBUFQB2673.1%61.7%56.0%67.3%
16Matthew StaffordLAQB54762.2%61.7%58.7%64.6%
17Andy Dalton◦ provisionalCARQB16863.1%61.6%57.2%65.8%
18Jalen HurtsPHIQB40062.0%61.4%58.0%64.7%
19Josh AllenBUFQB49861.7%61.2%58.1%64.3%
20Joshua Dobbs◦ provisionalSFQB4965.3%61.2%55.8%66.5%
21Brock PurdySFQB48761.6%61.2%58.1%64.3%
22Tyrod Taylor◦ provisionalNYJQB2568.0%61.1%55.3%66.7%
23Malik Willis◦ provisionalGBQB6363.5%61.0%55.7%66.1%
24Tanner McKee◦ provisionalPHIQB4763.8%60.9%55.4%66.2%
25Mac JonesJAXQB27961.3%60.8%57.0%64.6%
26Sam DarnoldMINQB59460.8%60.6%57.7%63.5%
27Mason RudolphTENQB24060.8%60.5%56.5%64.4%
28Tommy DeVito◦ provisionalNYGQB5062.0%60.5%55.1%65.8%
29Jordan LoveGBQB44260.6%60.5%57.2%63.7%
30Justin HerbertLACQB54860.6%60.5%57.4%63.4%
31Joe FlaccoINDQB26760.7%60.4%56.6%64.2%
32Gardner MinshewLVQB33660.4%60.3%56.7%63.8%
33Aidan O'ConnellLVQB25560.4%60.3%56.3%64.1%
34Jimmy Garoppolo◦ provisionalLAQB4560.0%60.0%54.5%65.4%
35Drake MayeNEQB37560.0%60.0%56.5%63.4%
36Dak PrescottDALQB30959.9%59.9%56.2%63.6%
37Justin FieldsPITQB17859.6%59.8%55.5%64.1%
38Trey Lance◦ provisionalDALQB4555.6%59.1%53.6%64.5%
39Aaron RodgersNYJQB62758.7%59.0%56.1%61.8%
40Brandon Allen◦ provisionalSFQB3253.1%59.0%53.3%64.5%
41Skylar Thompson◦ provisionalMIAQB3953.8%58.9%53.3%64.4%
42Kenny Pickett◦ provisionalPHIQB4654.4%58.9%53.3%64.3%
43Tyler Huntley◦ provisionalMIAQB15057.3%58.8%54.3%63.2%
44Cooper RushDALQB32258.1%58.8%55.1%62.4%
45Davis Mills◦ provisionalHOUQB3852.6%58.7%53.1%64.2%
46Russell WilsonPITQB37057.8%58.5%55.0%62.0%
47Bailey Zappe◦ provisionalCLEQB3250.0%58.5%52.8%64.1%
48Michael Penix Jr.◦ provisionalATLQB10956.0%58.5%53.6%63.2%
49Daniel JonesNYGQB37557.6%58.4%54.9%61.8%
50Tim Boyle◦ provisionalNYGQB5251.9%58.2%52.7%63.5%
51Desmond Ridder◦ provisionalLVQB9554.7%58.2%53.2%63.1%
52C.J. StroudHOUQB58457.5%58.1%55.1%61.0%
53Trevor LawrenceJAXQB30656.2%57.6%53.9%61.3%
54Bryce YoungCARQB41456.5%57.6%54.2%60.9%
55Drew LockNYGQB19455.1%57.5%53.2%61.7%
56Jameis WinstonCLEQB32755.4%57.0%53.3%60.6%
57Deshaun WatsonCLEQB25054.8%57.0%53.0%60.9%
58Will LevisTENQB34455.2%56.8%53.3%60.4%
59Jacoby BrissettNEQB18052.8%56.3%52.0%60.7%
60Caleb WilliamsCHIQB64054.8%56.0%53.1%58.8%
61Jake Haener◦ provisionalNOQB4540.0%55.9%50.4%61.4%
62Spencer RattlerNOQB25151.8%55.2%51.2%59.1%
63Dorian Thompson-Robinson◦ provisionalCLEQB12648.4%55.1%50.4%59.9%
64Anthony RichardsonINDQB27945.2%50.9%47.0%54.7%

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).