Bayesian NBA · overview

Posterior Probability

A web app that frames an NBA playoff series as a Bayesian inference problem. The per-game win probability has a Beta prior, conjugately updates after every game, and produces a credible interval that the sportsbook's de-vigged price either falls inside or outside. The four pages below render the four pieces of that argument on live data: a schedule-adjusted team-strength estimate, a series-by-series posterior tracker, the joint distribution of winner and series length, and a slate-level test against the market.

Every interval the app shows is paired with the assumption that produced it. The docs page collects the background, the math behind each module, and the things the model does not do.