Statistical Modeling for Sports

Overview

This skill's job is classical statistical inference on a user-owned sports table: choose a GLM or test, check assumptions, report effects and uncertainty.

It is not a universal modeling workflow. Do not use it to load data, design walk-forward ML horse races, or certify leakage. Mixed, Bayesian, causal, and power-analysis material below is routed methodology — the bundled scripts only cover univariate assumption checks and an additive numeric binomial GLM.

Conduct the analysis the way a careful analyst would: right model family for the outcome, verified assumptions, honest effect sizes and uncertainty, and a write-up that survives review.

Typical stack: pandas, numpy, scipy, statsmodels, and scikit-learn (+ optional pingouin, seaborn, pymc, and arviz).