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 % · 2021

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
216
attempts to trust the number
shrunkrawthe shrink90% CI
50%55%60%65%avg1Tom BradyTB · 74264.2%2Aaron RodgersGB · 56263.8%3Dak PrescottDAL · 63163.8%4Colt McCoyARI · 10563.6%5Kyler MurrayARI · 51463.4%6Derek CarrLV · 66963.0%7Joe BurrowCIN · 57362.9%8Mac JonesNE · 55162.8%9Matthew StaffordLA · 63462.8%10Tua TagovailoaMIA · 41062.8%11Jimmy GaroppoloSF · 47262.6%12Patrick MahomesKC · 68862.6%13Kirk CousinsMIN · 59262.1%14Justin HerbertLAC · 71061.9%15Mike WhiteNYJ · 13761.8%16Matt RyanATL · 60261.7%17Jared GoffDET · 53461.6%18Josh JohnsonBAL · 8861.5%19Tim BoyleDET · 9561.4%20Teddy BridgewaterDEN · 46061.4%21Ryan TannehillTEN · 57961.3%22Davis MillsHOU · 42661.2%23Nick MullensCLE · 3061.0%24Gardner MinshewPHI · 6660.7%25Josh AllenBUF · 67560.5%26Case KeenumCLE · 7760.4%27Joe FlaccoNYJ · 4460.4%28Daniel JonesNYG · 38460.3%29Garrett GilbertWAS · 3360.3%30Mason RudolphPIT · 5860.2%

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 % · 2021 · full board

Completion % leaderboard for the 2021 NFL season, shrunk by empirical Bayes with 90% credible intervals, ranked by the shrunk estimate.
#PlayerTeamPosattemptsRawShrunk90% interval
1Tom BradyTBQB74265.4%64.2%61.6%66.7%
2Aaron RodgersGBQB56265.1%63.8%60.9%66.6%
3Dak PrescottDALQB63165.0%63.8%61.0%66.5%
4Colt McCoy◦ provisionalARIQB10570.5%63.6%59.1%67.9%
5Kyler MurrayARIQB51464.8%63.4%60.5%66.3%
6Derek CarrLVQB66964.0%63.0%60.4%65.7%
7Joe BurrowCINQB57363.9%62.9%60.0%65.7%
8Mac JonesNEQB55163.9%62.8%60.0%65.7%
9Matthew StaffordLAQB63463.7%62.8%60.1%65.5%
10Tua TagovailoaMIAQB41064.1%62.8%59.6%65.9%
11Jimmy GaroppoloSFQB47263.8%62.6%59.6%65.7%
12Patrick MahomesKCQB68863.4%62.6%60.0%65.2%
13Kirk CousinsMINQB59262.8%62.1%59.3%64.9%
14Justin HerbertLACQB71062.4%61.9%59.2%64.5%
15Mike White◦ provisionalNYJQB13764.2%61.8%57.5%66.0%
16Matt RyanATLQB60262.3%61.7%58.9%64.5%
17Jared GoffDETQB53462.2%61.6%58.7%64.5%
18Josh Johnson◦ provisionalBALQB8864.8%61.5%56.9%66.1%
19Tim Boyle◦ provisionalDETQB9564.2%61.4%56.9%65.9%
20Teddy BridgewaterDENQB46062.0%61.4%58.3%64.5%
21Ryan TannehillTENQB57961.7%61.3%58.4%64.1%
22Davis MillsHOUQB42661.7%61.2%58.0%64.4%
23Nick Mullens◦ provisionalCLEQB3066.7%61.0%55.8%66.0%
24Gardner Minshew◦ provisionalPHIQB6662.1%60.7%55.8%65.4%
25Josh AllenBUFQB67560.6%60.5%57.8%63.2%
26Case Keenum◦ provisionalCLEQB7761.0%60.4%55.7%65.1%
27Joe Flacco◦ provisionalNYJQB4461.4%60.4%55.4%65.3%
28Daniel JonesNYGQB38460.4%60.3%57.0%63.6%
29Garrett Gilbert◦ provisionalWASQB3360.6%60.3%55.1%65.3%
30Mason Rudolph◦ provisionalPITQB5860.3%60.2%55.3%65.0%
31Ben RoethlisbergerPITQB64860.2%60.2%57.4%62.9%
32Geno Smith◦ provisionalSEAQB10860.2%60.2%55.7%64.6%
33Nick Foles◦ provisionalCHIQB4060.0%60.2%55.1%65.1%
34Cooper Rush◦ provisionalDALQB5060.0%60.2%55.2%65.0%
35Taylor HeinickeWASQB53560.0%60.1%57.1%63.0%
36Sean Mannion◦ provisionalMINQB3857.9%59.9%54.8%64.9%
37Russell WilsonSEAQB43559.5%59.8%56.6%62.9%
38Tyler Huntley◦ provisionalBALQB20858.7%59.4%55.5%63.3%
39Andy DaltonCHIQB25758.0%59.0%55.3%62.7%
40Jordan Love◦ provisionalGBQB6654.5%58.9%54.0%63.6%
41Carson WentzINDQB55258.3%58.9%55.9%61.8%
42Jacoby BrissettMIAQB24557.6%58.8%55.0%62.5%
43Trey Lance◦ provisionalSFQB7554.7%58.8%54.0%63.5%
44Lamar JacksonBALQB42557.9%58.7%55.5%61.8%
45Drew Lock◦ provisionalDENQB12055.8%58.6%54.2%63.0%
46Jalen HurtsPHIQB46157.5%58.4%55.2%61.5%
47Tyrod Taylor◦ provisionalHOUQB16355.8%58.3%54.1%62.5%
48Ian Book◦ provisionalNOQB2842.9%58.2%53.0%63.3%
49Jameis Winston◦ provisionalNOQB17255.2%58.0%53.8%62.1%
50Brandon Allen◦ provisionalCINQB3943.6%57.6%52.5%62.7%
51PJ Walker◦ provisionalCARQB7448.6%57.3%52.4%62.0%
52Trevor LawrenceJAXQB63956.2%57.2%54.4%60.0%
53Trevor Siemian◦ provisionalNOQB20153.7%57.1%53.1%61.0%
54Sam DarnoldCARQB44255.0%56.7%53.5%59.9%
55Cam Newton◦ provisionalCARQB13750.4%56.4%52.0%60.7%
56Baker MayfieldCLEQB46454.5%56.3%53.2%59.4%
57Mike Glennon◦ provisionalNYGQB17750.8%56.0%51.8%60.1%
58Jake Fromm◦ provisionalNYGQB6640.9%55.7%50.8%60.5%
59Justin FieldsCHIQB30652.0%55.4%51.8%58.9%
60Zach WilsonNYJQB43049.5%53.1%49.9%56.3%

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).