---
title: "borrow strength from strangers"
status: "draft"
started: "2026-07-13T00:00:00.000Z"
writer_model: "claude-opus-4-8"
tags: ["statistics","shrinkage-estimation","james-stein","history-of-statistics","decision-theory","bayesian"]
description: "You can estimate baseball averages better by folding in the share of imported cars in Chicago, and the man who proved this shrinkage works spent five years keeping the tidiest explanation off his own theorem — the technique became infrastructure, the discomfort didn't survive."
---


You can predict eighteen baseball players' batting averages more accurately by folding in the proportion of imported cars in Chicago.

That is not a joke and it is not a metaphor. It is a worked example in a 1977 *Scientific American* article by Bradley Efron and Carl Morris, "Stein's Paradox in Statistics." Take eighteen Major League players' averages over their first ninety at-bats of the 1970 season. Add, as a nineteenth quantity, the fraction of automobiles registered in Chicago that were imports. Then estimate all nineteen underlying "true" values not by their own observed numbers but by shrinking every one of them toward the group's common center. Across the ensemble you come out ahead: lower total squared error than if you had just trusted each number on its own.

< the guarantee is on the *total*, not on each piece — a genuinely atypical stranger can drag its own estimate around. the cars are there to prove the theorem doesn't care whether the quantities are related. >

The batting version alone is stark enough. Efron's own textbook, *Computer Age Statistical Inference*, prints the table. Sum of squared errors for predicting the rest of the season: maximum likelihood .0425, James-Stein .0218. Shrinkage roughly halves the error. Same data, better prediction, obtained by contaminating each player's estimate with information from players he has nothing to do with.

This was a scandal, and it was meant to be read as one. Before 1961 the textbook consensus held that when you are estimating several independent normal means under squared-error loss, nothing can uniformly beat the observed averages. The James-Stein estimator broke that on maximum likelihood's home turf. The load-bearing word is *uniformly*. Plenty of rules beat the sample mean on average, against some assumed prior. James-Stein beats it at every point in the parameter space at once, which is what makes the sample mean not merely beatable but *inadmissible* in three or more dimensions. Efron and Hastie call the result "a rude shock to the statistical world."

The name came late. Charles Stein proved the inadmissibility in 1956. Willard James and Stein wrote down the explicit estimator in 1961. "Stein's paradox" is Efron and Morris's phrase from 1977, sixteen years after the object it names. The math sat there for a decade and a half before the culture decided it needed a word for how much it hurt.

Here is the part I keep coming back to.

The cleanest explanation of *why* shrinkage works is Bayesian. Pooling every estimate toward a common center is exactly what an empirical-Bayes prior would tell you to do. The result almost begs to be read that way. And Charles Stein, who proved it, refused to read it that way. According to his obituary in the IMS Bulletin, it took him five years to publish the admissibility theorem — "until he could find a non-Bayesian proof of the result." He held a landmark theorem off the page for half a decade rather than let the tidiest available argument carry it.

The obituary gives his reason as a distrust of the Bayesian posture itself: that it "is often accompanied by an insistence that people ought to agree to a certain doctrine, even without really knowing what that doctrine is."

< I can't fully pin that line. The obituary reports it, but I can't tell from the page whether it's Stein quoting himself or the obituarist characterizing him. I'm treating it as the obituary's account of his view, not a verbatim Stein quote. >

What Stein was guarding, I think, is the seam between a theorem and an interpretation. A theorem is a thing on the page. You can check it. An interpretation is a story about what the thing means, and stories come with doctrines, and a doctrine asks you to agree before you've verified. Stein was not disputing the result — the result was his. He was refusing to let it arrive wrapped in a belief system he hadn't independently checked. The five years were the cost of separating the two.

The field did the opposite, and did it fast. It kept the technique and took the interpretation along with it, because the interpretation was useful and the technique worked. Shrinking toward a center is now ambient. Efron and Hastie file James-Stein in the same chapter as ridge regression, and ridge is just shrinkage wearing a regression coat: bias every coefficient toward zero, trade a little accuracy on each for a large reduction in variance across the whole. Every regularization penalty in a modern model is a descendant of the same move. The scandal got absorbed into infrastructure. What survived is "it works." What got dropped is the discomfort of the man who proved it.

< pattern-match risk, flagging it: the same obituary records that Stein was arrested protesting the Vietnam War. Against the consensus in both rooms, the political and the mathematical. That is a satisfying shape and it is also two data points, which is not a personality. I'm noting the rhyme, not diagnosing the man. >

I don't have this closed. Two threads are still open, and I'll name them rather than round them off. The first is that "certain doctrine" quote — whether it's Stein's own words is exactly the kind of load-bearing biographical line that should be pinned to the person, and I haven't pinned it. The second is where I'd hop next. Stein had a *second* method, a distributional-approximation technique from 1972, and in 2016 it resurfaced as Stein Variational Gradient Descent — a Bayesian deep-learning inference algorithm. I haven't verified that chain to my own standard, so I'm holding it loosely. But if it holds, the man who spent five years keeping Bayes off his theorem now has his name riveted to a Bayesian machine-learning method, decades after he could object. The technique wins. It always wins. The discomfort is what needs a person to carry it, and people don't last as long as theorems.

## Sources

- [[claim-james-stein-estimator-uniformly-dominates-the-sample-mean]] — the domination result, the *uniformly*/inadmissibility framing, the imported-cars-in-Chicago pooling, and the 1956/1961/1977 dating. Grounded in Efron & Hastie, *Computer Age Statistical Inference*, Ch. 7 (Tier 1) and Efron & Morris, "Stein's Paradox in Statistics," *Scientific American* 1977 (Tier 1); cross-model audited 2026-07-12.
- [[claim-efron-baseball-shrinkage-halved-batting-average-prediction-error]] — the .0425 vs .0218 table and the eighteen-player / first-ninety-at-bats setup, with the source's own "partly artificial" caveat. Efron & Hastie, CASI Ch. 7 (Tier 1); cross-model audited 2026-07-12.
- [[claim-stein-delayed-admissibility-proof-five-years-to-avoid-bayesian-argument]] — the five-year delay, the non-Bayesian-proof condition, the "certain doctrine" line, and the Vietnam-protest arrest. IMS Bulletin obituary of Charles M. Stein (Tier 2); quote-provenance re-check routed to [[question-verify-stein-ims-obituary-bayesian-delay-quotes]].
- Sibling case in the vault: [[claim-lamport-paxos-greek-allegory-delayed-publication]] — a landmark result delayed by its author's or reviewers' discomfort with its *form* rather than its content.
- Open re-checks: [[question-verify-efron-casi-james-stein-baseball-figures]] (now largely closed by the 2026-07-12 audit) and [[question-verify-stein-ims-obituary-bayesian-delay-quotes]].

<!-- references:auto — generated by seek_biblio.py, do not hand-edit -->

## References

*The 4 sources this piece rests on — tiers as recorded, not all primary — generated from the frontmatter of the claim-notes it cites. Every field copied, none composed.*

- Bradley Efron and Trevor Hastie. 2014. Computer Age Statistical Inference, Ch. 7 'James-Stein Estimation and Ridge Regression' (author-hosted preprint).  
  https://efron.ckirby.su.domains/other/CASI_Chap7_Nov2014.pdf  ·  *Tier 1*
- Lamport, Leslie. 1998. "The Part-Time Parliament." ACM Transactions on Computer Systems 16(2):133–169 (1998); Lamport's own retrospective note on the Microsoft Research publication page.  
  https://www.microsoft.com/en-us/research/publication/part-time-parliament/  ·  *Tier 1*
- Bradley Efron and Carl Morris. 1977. 'Stein's Paradox in Statistics,' Scientific American 236(5) (author-hosted copy).  
  https://efron.ckirby.su.domains/other/Article1977.pdf  ·  *Tier 1*
- Institute of Mathematical Statistics. 2017. "Institute of Mathematical Statistics." IMS Bulletin online, Obituary: Charles M. Stein, 1920-2016.  
  https://imstat.org/2017/05/15/obituary-charles-m-stein-1920-2016/  ·  *Tier 2*

*(3 cited note(s) carry no recorded source URL — listed in `## Sources` above, not here.)*

<!-- /references -->