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 · 2022

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
163
dropbacks to trust the number
shrunkrawthe shrink90% CI
35%40%45%50%avg1Patrick MahomesKC · 67750.9%2Trevor LawrenceJAX · 61548.7%3Dak PrescottDAL · 41448.7%4Joe BurrowCIN · 65148.4%5Josh AllenBUF · 60048.2%6Jimmy GaroppoloSF · 32648.0%7Tua TagovailoaMIA · 42147.7%8Andy DaltonNO · 40447.6%9Jared GoffDET · 61047.6%10Geno SmithSEA · 61947.1%11Tom BradyTB · 75846.9%12Kirk CousinsMIN · 69046.9%13Nick MullensMIN · 2546.7%14Brock PurdySF · 18146.4%15Matthew StaffordLA · 33245.8%16Daniel JonesNYG · 51445.6%17Bailey ZappeNE · 9845.5%18Marcus MariotaATL · 32645.1%19Justin HerbertLAC · 73645.1%20Teddy BridgewaterMIA · 8745.1%21Sam DarnoldCAR · 14745.1%22Ryan TannehillTEN · 35845.1%23Jameis WinstonNO · 12545.1%24Jalen HurtsPHI · 49944.7%25Jarrett StidhamLV · 9044.6%26Lamar JacksonBAL · 35244.4%27Jacoby BrissettCLE · 39244.3%28Aaron RodgersGB · 57544.3%29Mitchell TrubiskyPIT · 19144.0%30Taylor HeinickeWAS · 27643.9%

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 · 2022 · full board

Dropback success rate leaderboard for the 2022 NFL season, shrunk by empirical Bayes with 90% credible intervals, ranked by the shrunk estimate.
#PlayerTeamPosdropbacksRawShrunk90% interval
1Patrick MahomesKCQB67752.6%50.9%48.1%53.8%
2Trevor LawrenceJAXQB61549.9%48.7%45.7%51.6%
3Dak PrescottDALQB41450.5%48.7%45.2%52.1%
4Joe BurrowCINQB65149.5%48.4%45.5%51.2%
5Josh AllenBUFQB60049.3%48.2%45.2%51.2%
6Jimmy GaroppoloSFQB32650.0%48.0%44.3%51.7%
7Tua TagovailoaMIAQB42149.2%47.7%44.3%51.1%
8Andy DaltonNOQB40449.0%47.6%44.1%51.0%
9Jared GoffDETQB61048.5%47.6%44.6%50.5%
10Geno SmithSEAQB61948.0%47.1%44.2%50.1%
11Tom BradyTBQB75847.5%46.9%44.2%49.6%
12Kirk CousinsMINQB69047.5%46.9%44.1%49.7%
13Nick Mullens◦ provisionalMINQB2564.0%46.7%40.7%52.7%
14Brock PurdySFQB18148.6%46.4%42.0%50.9%
15Matthew StaffordLAQB33246.7%45.8%42.1%49.5%
16Daniel JonesNYGQB51446.1%45.6%42.5%48.8%
17Bailey Zappe◦ provisionalNEQB9848.0%45.5%40.5%50.6%
18Marcus MariotaATLQB32645.7%45.1%41.4%48.9%
19Justin HerbertLACQB73645.4%45.1%42.4%47.9%
20Teddy Bridgewater◦ provisionalMIAQB8747.1%45.1%40.0%50.3%
21Sam Darnold◦ provisionalCARQB14746.3%45.1%40.5%49.7%
22Ryan TannehillTENQB35845.5%45.1%41.5%48.6%
23Jameis Winston◦ provisionalNOQB12546.4%45.1%40.3%49.9%
24Jalen HurtsPHIQB49944.9%44.7%41.5%47.9%
25Jarrett Stidham◦ provisionalLVQB9045.6%44.6%39.5%49.7%
26Lamar JacksonBALQB35244.6%44.4%40.8%48.0%
27Jacoby BrissettCLEQB39244.4%44.3%40.8%47.8%
28Aaron RodgersGBQB57544.4%44.3%41.3%47.3%
29Mitchell TrubiskyPITQB19144.0%44.0%39.7%48.4%
30Taylor HeinickeWASQB27643.8%43.9%40.0%47.8%
31John Wolford◦ provisionalLAQB6943.5%43.9%38.5%49.2%
32Gardner Minshew◦ provisionalPHIQB8143.2%43.7%38.6%49.0%
33Matt RyanINDQB50243.4%43.6%40.4%46.7%
34Tyler Huntley◦ provisionalBALQB11742.7%43.5%38.6%48.4%
35Kenny PickettPITQB41743.2%43.4%40.0%46.8%
36Mike WhiteNYJQB18442.4%43.1%38.8%47.5%
37Trevor Siemian◦ provisionalCHIQB2835.7%42.8%37.0%48.7%
38Desmond Ridder◦ provisionalATLQB12441.1%42.8%38.0%47.6%
39Joshua Dobbs◦ provisionalTENQB7439.2%42.5%37.3%47.8%
40Nathan Peterman◦ provisionalCHIQB2532.0%42.4%36.5%48.4%
41Deshaun WatsonCLEQB18841.0%42.4%38.1%46.7%
42David Blough◦ provisionalARIQB6338.1%42.4%37.0%47.8%
43Kyler MurrayARIQB41841.6%42.3%38.9%45.7%
44Trey Lance◦ provisionalSFQB3333.3%42.2%36.5%48.0%
45Davis Webb◦ provisionalNYGQB4134.2%42.0%36.4%47.7%
46Derek CarrLVQB53241.3%42.0%38.9%45.1%
47Brett Rypien◦ provisionalDENQB9638.5%42.0%37.0%47.0%
48Cooper RushDALQB16839.3%41.6%37.2%46.1%
49Colt McCoy◦ provisionalARIQB14538.6%41.5%36.9%46.1%
50Joe FlaccoNYJQB20139.3%41.4%37.2%45.7%
51Nick Foles◦ provisionalINDQB5032.0%41.2%35.7%46.8%
52Bryce Perkins◦ provisionalLAQB3928.2%40.9%35.3%46.7%
53Carson WentzWASQB30938.8%40.6%36.9%44.4%
54PJ Walker◦ provisionalCARQB11335.4%40.5%35.7%45.4%
55Mac JonesNEQB47839.1%40.4%37.2%43.6%
56Kyle Allen◦ provisionalHOUQB8532.9%40.2%35.1%45.4%
57Skylar Thompson◦ provisionalMIAQB11134.2%40.1%35.2%45.0%
58Zach WilsonNYJQB26537.4%39.9%36.0%43.8%
59Anthony Brown◦ provisionalBALQB5226.9%39.9%34.4%45.4%
60Sam Ehlinger◦ provisionalINDQB11533.9%39.8%35.0%44.7%
61Davis MillsHOUQB51137.4%39.0%35.9%42.1%
62Trace McSorley◦ provisionalARIQB8629.1%38.8%33.8%44.0%
63Justin FieldsCHIQB37736.6%38.8%35.4%42.3%
64Malik Willis◦ provisionalTENQB7126.8%38.8%33.6%44.1%
65Russell WilsonDENQB53836.6%38.3%35.3%41.4%
66Baker MayfieldLAQB37135.6%38.1%34.7%41.6%

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).