Module mc_stats 

Streaming moments and Bernoulli intervals for Monte Carlo hit statistics.

Welford accumulates a running mean and variance in constant memory, one trial at a time, so an adaptive Monte Carlo driver never needs to retain the full trial history just to report a standard deviation. wilson_interval is the classic fixed-n confidence interval for a Bernoulli hit/miss proportion, evaluated at one of three pinned confidence levels (ConfidenceLevel). BernoulliConfidenceSequence is the anytime-valid beta-binomial-mixture counterpart to wilson_interval: it lets an adaptive driver peek at partial results after every trial, and stop on a data-dependent rule, without inflating its error rate (Task 3 of Plan C, MBA-1352).

No randomness lives in the production path: this module consumes trial outcomes a caller already produced (MonteCarloTrialSampler::sample_one_trial in src/cli_api.rs, the one trial body both the legacy fixed-count loop and the adaptive driver run) and is pure std math only – no rand, no fs, no clap – so it compiles for wasm32-unknown-unknown unconditionally. (One #[cfg(test)] test, the empirical coverage check, does draw from a seeded rand generator; nothing outside #[cfg(test)] does.)

Structs

BernoulliConfidenceSequence

Anytime-valid confidence sequence for a Bernoulli proportion: Robbins’ beta-binomial mixture with a uniform Beta(1, 1) prior.

Welford

Online (streaming) mean and variance via Welford’s algorithm.

Enums

ConfidenceLevel

A confidence level pinned to one of three fixed two-sided critical values.

Functions

wilson_interval

Wilson score interval for a Bernoulli proportion at fixed n.