Bellz Theorem, bonus: Stop When You Know
Abraham Wald pays the Labs a visit. Gus checks a crate of relays the thorough way, row by row down a fixed sampling table; Wald tests one at a time and stops the moment the evidence is enough, then goes home. They reach the same answer. On screen and true: Shewhart's control chart (1924) and Dodge and Romig's sampling tables (1929 on) at the Labs; Wald's sequential analysis, worked out at Columbia's Statistical Research Group in 1943 and kept secret until 1945, and Wald's own estimate that it needs about half the tests, on average. Labelled liberties: Wald never worked at Bell Labs; the parts crate (his question came from ordnance testing); stopping at item seven. The armor sketch at the end is a later retelling of Wald's 1943 memos. It closes on footage of Where the Holes Aren't, playable on the Games row. Music made for the film (ACE-Step).
True story Every fact on screen was checked before release (2026-10-29). The dialogue is invented.
Sources
- NIST/SEMATECH e-Handbook: Shewhart's control chart memo, 16 May 1924
- Dodge and Romig, A Method of Sampling Inspection, BSTJ 8 (1929)
- Wallis, The Statistical Research Group, 1942-1945, JASA 75 (1980)
- Wald, Sequential Tests of Statistical Hypotheses, Annals of Math. Stat. 16 (1945)
- Wald, Sequential Analysis (Wiley, 1947), p. 1: about 50 per cent fewer observations on average
- Mangel and Samaniego, Abraham Wald's Work on Aircraft Survivability, JASA 79 (1984)
