Bayesian modelling of shot generation and conversion in soccer.
The model is described briefly in a blogpost, for full definition consult the shotgen.stan and shotconv.stan files.
Currently the model is best executed and visualised via IPython, like so:
%run run.py
gen_samples, team_map = run("shotgen.stan")
conv_samples, _ = run("shotconv.stan")
%run analysis.py
team_profiles(gen_samples, conv_samples, team_map)
score_bars(gen_samples)
score_bars(conv_samples)which produces figures like in the blogpost.