|
Abstract:
|
The advent of high-resolution pitch tracking data (PITCHf/x) has facilitated many quantitative analyses of "pitch framing." Framing refers to the ability of Major League Baseball catchers to catch a pitch in such a way as to increase the chance that the umpire calls the pitch a strike. Multiple analyses, utilizing a range of modeling techniques, all suggest that framing can have an outsize effect, with a good framer able to save his team anywhere on the order of 20 - 50 runs over the course of a season.
In this talk, I will revisit one such analysis based on fitting a hierarchical Bayesian model that partially pooled between umpires. I will discuss some new refinements to this analysis, focusing, in particular, on new models for the called strike probabilities and the value of a called strike based on Bayesian additive regression trees
|
