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 % · 2020

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
223
attempts to trust the number
shrunkrawthe shrink90% CI
55%60%65%avg1Aaron RodgersGB · 54666.1%2Drew BreesNO · 40465.6%3Josh AllenBUF · 59964.7%4Philip RiversIND · 56464.2%5Teddy BridgewaterCAR · 52563.7%6Jared GoffLA · 57663.3%7Deshaun WatsonHOU · 59763.2%8Derek CarrLV · 54463.1%9Ben RoethlisbergerPIT · 62563.1%10Kyler MurrayARI · 58763.1%11Ryan FitzpatrickMIA · 28363.1%12Patrick MahomesKC · 61362.9%13Dak PrescottDAL · 23462.8%14Tom BradyTB · 63362.7%15Russell WilsonSEA · 60762.7%16Chad HenneKC · 4062.4%17Justin HerbertLAC · 63062.4%18Mitchell TrubiskyCHI · 31662.2%19Kirk CousinsMIN · 56261.8%20Brett RypienDEN · 4161.8%21Ryan TannehillTEN · 50761.8%22Jimmy GaroppoloSF · 15161.5%23Kyle AllenWAS · 9661.5%24Chase DanielDET · 4661.4%25Alex SmithWAS · 27461.2%26Nick FolesCHI · 33161.0%27Marcus MariotaLV · 2861.0%28Nick MullensSF · 34760.9%29Matt RyanATL · 67060.8%30Brandon AllenCIN · 14960.8%31Joe BurrowCIN · 43660.7%

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 % · 2020 · full board

Completion % leaderboard for the 2020 NFL season, shrunk by empirical Bayes with 90% credible intervals, ranked by the shrunk estimate.
#PlayerTeamPosattemptsRawShrunk90% interval
1Aaron RodgersGBQB54668.1%66.1%63.2%68.9%
2Drew BreesNOQB40468.1%65.6%62.4%68.7%
3Josh AllenBUFQB59966.1%64.7%62.0%67.5%
4Philip RiversINDQB56465.4%64.2%61.4%67.0%
5Teddy BridgewaterCARQB52564.8%63.7%60.8%66.5%
6Jared GoffLAQB57664.2%63.3%60.5%66.1%
7Deshaun WatsonHOUQB59764.0%63.2%60.4%65.9%
8Derek CarrLVQB54464.0%63.1%60.2%66.0%
9Ben RoethlisbergerPITQB62563.8%63.1%60.4%65.8%
10Kyler MurrayARIQB58763.9%63.1%60.3%65.9%
11Ryan FitzpatrickMIAQB28364.7%63.1%59.5%66.6%
12Patrick MahomesKCQB61363.6%62.9%60.2%65.7%
13Dak PrescottDALQB23464.5%62.8%59.1%66.5%
14Tom BradyTBQB63363.3%62.7%60.0%65.5%
15Russell WilsonSEAQB60763.3%62.7%59.9%65.4%
16Chad Henne◦ provisionalKCQB4070.0%62.4%57.5%67.3%
17Justin HerbertLACQB63062.9%62.4%59.7%65.1%
18Mitchell TrubiskyCHIQB31663.0%62.2%58.7%65.6%
19Kirk CousinsMINQB56262.1%61.8%58.9%64.6%
20Brett Rypien◦ provisionalDENQB4165.8%61.8%56.9%66.7%
21Ryan TannehillTENQB50762.1%61.8%58.8%64.7%
22Jimmy Garoppolo◦ provisionalSFQB15162.3%61.5%57.4%65.6%
23Kyle Allen◦ provisionalWASQB9662.5%61.5%57.0%65.9%
24Chase Daniel◦ provisionalDETQB4663.0%61.4%56.5%66.2%
25Alex SmithWASQB27461.3%61.2%57.6%64.8%
26Nick FolesCHIQB33161.0%61.0%57.6%64.4%
27Marcus Mariota◦ provisionalLVQB2860.7%61.0%55.9%66.0%
28Nick MullensSFQB34760.8%60.9%57.5%64.2%
29Matt RyanATLQB67060.8%60.8%58.1%63.5%
30Brandon Allen◦ provisionalCINQB14960.4%60.8%56.6%64.9%
31Joe BurrowCINQB43660.6%60.7%57.6%63.8%
32Andy DaltonDALQB35760.5%60.7%57.4%64.0%
33Brian Hoyer◦ provisionalNEQB2657.7%60.7%55.6%65.7%
34Cam NewtonNEQB40060.5%60.7%57.5%63.9%
35Gardner MinshewJAXQB35860.3%60.6%57.3%63.9%
36John Wolford◦ provisionalLAQB4055.0%60.2%55.1%65.0%
37Mason Rudolph◦ provisionalPITQB4555.6%60.1%55.2%65.0%
38Lamar JacksonBALQB40659.6%60.1%56.9%63.3%
39Tua TagovailoaMIAQB31359.4%60.1%56.6%63.6%
40C.J. Beathard◦ provisionalSFQB11457.9%60.0%55.6%64.3%
41Baker MayfieldCLEQB51359.5%59.9%57.0%62.9%
42Matthew StaffordDETQB57059.5%59.9%57.0%62.8%
43Garrett Gilbert◦ provisionalDALQB4052.5%59.8%54.8%64.7%
44Colt McCoy◦ provisionalNYGQB7255.6%59.7%55.0%64.4%
45Mike Glennon◦ provisionalJAXQB19158.1%59.7%55.7%63.6%
46Tyrod Taylor◦ provisionalLACQB3250.0%59.7%54.6%64.7%
47PJ Walker◦ provisionalCARQB6053.3%59.4%54.6%64.2%
48Dwayne HaskinsWASQB26256.5%58.6%54.9%62.3%
49Jarrett Stidham◦ provisionalNEQB4845.8%58.4%53.4%63.2%
50Ben DiNucci◦ provisionalDALQB5046.0%58.3%53.4%63.2%
51Ryan Finley◦ provisionalCINQB4141.5%58.0%53.0%63.0%
52Daniel JonesNYGQB49656.5%57.9%54.9%60.9%
53Joe Flacco◦ provisionalNYJQB14152.5%57.8%53.5%62.0%
54Jake Luton◦ provisionalJAXQB11850.8%57.5%53.1%61.9%
55Jeff Driskel◦ provisionalDENQB7646.1%57.3%52.5%61.9%
56Drew LockDENQB46654.5%56.6%53.5%59.7%
57Sam DarnoldNYJQB40154.1%56.6%53.3%59.9%
58Jalen Hurts◦ provisionalPHIQB16247.5%55.4%51.2%59.5%
59Carson WentzPHIQB49350.9%54.1%51.0%57.1%

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).