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 · 2018

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
192
dropbacks to trust the number
shrunkrawthe shrink90% CI
35%40%45%50%55%avg1Drew BreesNO · 50852.9%2Patrick MahomesKC · 60650.3%3Tom BradyNE · 59049.8%4Jared GoffLA · 59549.7%5Philip RiversLAC · 54749.7%6Andrew LuckIND · 65849.4%7Matt RyanATL · 65549.4%8Ben RoethlisbergerPIT · 70148.9%9Carson WentzPHI · 43148.6%10Ryan FitzpatrickTB · 26148.6%11Jameis WinstonTB · 40747.9%12Cam NewtonCAR · 50147.9%13Kirk CousinsMIN · 64847.8%14Derek CarrLV · 59847.2%15Mitchell TrubiskyCHI · 46147.2%16Jimmy GaroppoloSF · 10347.1%17Deshaun WatsonHOU · 56247.1%18Andy DaltonCIN · 38446.9%19Matt BarkleyBUF · 2646.8%20Nick MullensSF · 28846.5%21Kyle AllenCAR · 3146.2%22Lamar JacksonBAL · 18645.8%23Chase DanielCHI · 8645.7%24Russell WilsonSEA · 47145.7%25Colt McCoyWAS · 5845.6%26C.J. BeathardSF · 18945.4%27Nick FolesPHI · 20345.3%28Baker MayfieldCLE · 51545.3%29Aaron RodgersGB · 64745.2%30Marcus MariotaTEN · 37445.1%

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 · 2018 · full board

Dropback success rate leaderboard for the 2018 NFL season, shrunk by empirical Bayes with 90% credible intervals, ranked by the shrunk estimate.
#PlayerTeamPosdropbacksRawShrunk90% interval
1Drew BreesNOQB50855.5%52.9%49.8%56.0%
2Patrick MahomesKCQB60651.6%50.3%47.3%53.2%
3Tom BradyNEQB59051.0%49.8%46.8%52.7%
4Jared GoffLAQB59550.9%49.7%46.8%52.6%
5Philip RiversLACQB54751.0%49.7%46.7%52.7%
6Andrew LuckINDQB65850.5%49.4%46.6%52.2%
7Matt RyanATLQB65550.4%49.4%46.5%52.2%
8Ben RoethlisbergerPITQB70149.8%48.9%46.2%51.7%
9Carson WentzPHIQB43149.9%48.6%45.4%51.9%
10Ryan FitzpatrickTBQB26150.6%48.6%44.7%52.4%
11Jameis WinstonTBQB40748.9%47.9%44.6%51.3%
12Cam NewtonCARQB50148.7%47.9%44.8%51.0%
13Kirk CousinsMINQB64848.3%47.8%44.9%50.6%
14Derek CarrLVQB59847.7%47.2%44.3%50.1%
15Mitchell TrubiskyCHIQB46147.7%47.2%44.0%50.4%
16Jimmy Garoppolo◦ provisionalSFQB10349.5%47.1%42.4%51.9%
17Deshaun WatsonHOUQB56247.5%47.1%44.1%50.1%
18Andy DaltonCINQB38447.4%46.9%43.5%50.3%
19Matt Barkley◦ provisionalBUFQB2653.8%46.8%41.3%52.4%
20Nick MullensSFQB28846.9%46.5%42.7%50.2%
21Kyle Allen◦ provisionalCARQB3148.4%46.2%40.8%51.7%
22Lamar Jackson◦ provisionalBALQB18645.7%45.8%41.6%50.0%
23Chase Daniel◦ provisionalCHIQB8645.4%45.7%40.8%50.6%
24Russell WilsonSEAQB47145.6%45.7%42.5%48.9%
25Colt McCoy◦ provisionalWASQB5844.8%45.6%40.5%50.8%
26C.J. Beathard◦ provisionalSFQB18945.0%45.4%41.3%49.6%
27Nick FolesPHIQB20344.8%45.3%41.2%49.5%
28Baker MayfieldCLEQB51545.1%45.3%42.2%48.4%
29Aaron RodgersGBQB64745.0%45.2%42.4%48.0%
30Marcus MariotaTENQB37444.6%45.1%41.6%48.5%
31Joe FlaccoBALQB39444.4%44.9%41.5%48.3%
32Teddy Bridgewater◦ provisionalNOQB2536.0%44.7%39.2%50.3%
33Dak PrescottDALQB58344.3%44.6%41.7%47.6%
34Josh Johnson◦ provisionalWASQB10242.2%44.6%39.8%49.4%
35Alex SmithWASQB34943.8%44.6%41.1%48.1%
36Cody Kessler◦ provisionalJAXQB15341.8%44.1%39.7%48.5%
37Taylor Heinicke◦ provisionalCARQB5837.9%44.0%38.9%49.2%
38Blaine Gabbert◦ provisionalTENQB10640.6%44.0%39.3%48.7%
39Matthew StaffordDETQB59742.9%43.6%40.7%46.5%
40Eli ManningNYGQB62742.7%43.5%40.6%46.3%
41Case KeenumDENQB62042.7%43.5%40.6%46.4%
42Derek Anderson◦ provisionalBUFQB7537.3%43.5%38.5%48.5%
43Brock OsweilerMIAQB19540.5%43.2%39.1%47.3%
44Sam DarnoldNYJQB44441.7%42.9%39.7%46.2%
45Mark Sanchez◦ provisionalWASQB4228.6%42.8%37.5%48.1%
46Josh McCown◦ provisionalNYJQB11637.1%42.6%38.0%47.2%
47Blake BortlesJAXQB43641.1%42.5%39.3%45.8%
48Ryan TannehillMIAQB30839.9%42.2%38.6%45.9%
49Jeff DriskelCINQB19538.0%41.9%37.8%46.0%
50DeShone Kizer◦ provisionalGBQB4524.4%41.8%36.6%47.1%
51Sam Bradford◦ provisionalARIQB8731.0%41.2%36.4%46.1%
52Tyrod Taylor◦ provisionalCLEQB9730.9%40.8%36.1%45.6%
53Josh AllenBUFQB34837.9%40.8%37.3%44.3%
54Nathan Peterman◦ provisionalBUFQB8927.0%39.9%35.1%44.7%
55Josh RosenARIQB43835.4%38.6%35.4%41.8%

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).