A follow-up exploration of computing posterior distributions for RPG dice rolls conditioned on unlikely survival scenarios. The author builds a performant Python DSL using numpy that avoids brute-force Monte Carlo inefficiency through three key techniques: a backward liveness pass to eliminate unnecessary state and prevent
Table of contents
Biased posteriors the hard wayBiased posteriors an easier wayAuto batchingLet's work in a general semiringNot just PythonSort: