SEAM methodology for context-rich player matchup evaluations

TL;DR

The paper introduces the SEAM methodology, a fast, distributional approach for evaluating batter-pitcher matchups in baseball, aiding strategic decisions through a web app that visualizes spray chart distributions.

Contribution

The paper presents a novel synthetic matchup estimation method that improves reliability of spray chart predictions and provides a fast, web-based visualization tool for baseball strategy analysis.

Findings

Method accurately estimates spray chart distributions.

Web app enables instant visualization of matchups.

Synthetic players reduce small sample size issues.

Abstract

We develop the SEAM (synthetic estimated average matchup) method for describing batter versus pitcher matchups in baseball. We first estimate the distribution of balls put into play by a batter facing a pitcher, called the empirical spray chart distribution. Many individual matchups have a sample size that is too small to be reliable for use in predicting future outcomes. Synthetic versions of the batter and pitcher under consideration are constructed in order to alleviate these concerns. Weights governing how much influence these synthetic players have on the overall estimated spray chart distribution are constructed to minimize expected mean square error. We provide a Shiny web application that allows users to visualize and evaluate any batter-pitcher matchup that has occurred or could have occurred during the Statcast era (specifically 2017-present). This methodology and web application could be used to determine defensive alignments, lineup construction, or pitcher selection through estimation of spray densities based on any input matchup. One can access this web application at https://seam.stat.illinois.edu/. The computational speed with which the method calculates the spray densities allows the app to display the visualizations for any input almost instantly. Therefore, SEAM offers distributional interpretations of dependent matchup data which is computationally fast.