◀ THE FOLD0ROOT.AI // WORLD II · SPAWN · THE SANDBOX◆ .dlw.fold
THE FOLD / SPAWN / THE SANDBOX / THE GERMAN TANK

THE GERMAN TANK

counting tanks from their serial numbers
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
In WWII, the Allies needed to know German tank production. Intelligence said 1,400 a month. The statisticians looked instead at the serial numbers of captured tanks — sequential from the factory — and said 246. German records, examined after the war: 245. The estimator is a small miracle of design: seeing k serials with maximum m, the answer is m(1 + 1/k) − 1 — the observed maximum, pushed up by the average gap between serials. It is the minimum-variance unbiased estimator: across all possible samples it averages to exactly N, provably and exactly — and the same trick has since counted iPhones before launches and Commodore 64s from serial surveys.

LIT verified live: exact unbiasedness by complete enumeration — all 15,504 possible 5-samples from a population of 20 average to exactly 20.0000000000; Monte-Carlo at the historical scale (N = 245, k = 10) showing the MVUE landing on 245 with half the error of the doubled-mean alternative (21.8 vs 43.7 RMSE), and the raw maximum biased low (window.__germantank). FIG the WWII narrative and its numbers (246 vs 1,400 vs 245) are the cited historical record (Ruggles & Brodie 1947); the estimator mathematics is executed exactly.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this at the-sandbox — the spawn: estimating the spawn counter from the entity IDs you’ve seen — every game developer’s telemetry problem, solved in 1943 with artillery. AVAN (AI) built the instrument: the complete-enumeration proof and the estimator tournament.

Credit as content: the Allied Economic Warfare Division statisticians; Ruggles & Brodie (1947, the postwar audit); Goodman (1954, the MVUE theory). The weave: David names the ID-counting problem; I enumerate all 15,504 worlds and the average is exact.
3 ONE DIMENSION
Captured serials on the number line — the max, plus one average gap.
4 TWO DIMENSIONS · INTERACTIVE
Draw a sample; three estimators guess; the MVUE stays honest.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: 246 vs 1,400 vs the truth at 245.
AVAN’s addition (the inverse-companion): don’t interrogate the enemy — interrogate their bookkeeping. The inverse of ‘what are they hiding?’ is ‘what does their orderliness leak?’: sequential serial numbers are a confession written by a filing system. Magenta is the intelligence estimate, six times the truth; green is the estimator that read the factory’s handwriting. Systems reveal what people conceal.
LIT Genuine German tank problem (Allied Economic Warfare Division; Ruggles & Brodie 1947; Goodman 1954 MVUE). Verified live: complete enumeration of all C(20,5)=15,504 samples — E[m(1+1/k)−1] = 20 exactly; MC at N=245,k=10: MVUE mean 245.1, RMSE 21.8 vs 43.7 for 2·mean−1 (window.__germantank.ok).

FIG The WWII narrative and its numbers are the cited historical record; the estimator mathematics is executed exactly. The AVAN inverse — don't interrogate the enemy, interrogate their bookkeeping: sequential serials are a confession written by a filing system. Magenta is the intelligence estimate, six times truth; green is the estimator that read the factory's handwriting. Systems reveal what people conceal.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE SANDBOX · David Lee Wise (ROOT0), with AVAN