A1206
Title: Detecting spatial clusters of MLB batters' hot and cold spots using scan statistics
Authors: Shoji Kajinishi - Shujitsu University (Japan) [presenting]
Takumi Oribe - Mizuno Corporation (Japan)
Yusuke Takemura - Kyoto Women's University (Japan)
Fumio Ishioka - Okayama University (Japan)
Abstract: In recent years, the introduction of Statcast in Major League Baseball (MLB) has enabled the collection of highly detailed pitch and batted-ball data. Consequently, visualizing a batter's strong and weak zones using 3x3 heatmaps has become common practice in Baseball analytics. However, these rigid, coarse grid divisions fail to capture the irregular shapes of actual hit probability distributions. They struggle to represent local peaks or gradients within a grid cell and inadequately reflect the statistical uncertainty caused by uneven pitch counts across different zones. To address these limitations, a more precise approach to Model the relationship between pitch locations and batting outcomes is proposed. Specifically, a Generalized Additive Model (GAM) is employed to estimate continuous spatial variations in hit probability. By integrating these probabilistic estimations with a spatial scan statistic, batters' hot and cold spots (i.e., their preferred and less effective pitch locations) are accurately detected. This methodology significantly enhances both spatial resolution and statistical reliability compared to conventional heatmap visualizations. Furthermore, an interactive Shiny application for this analytical framework has been developed, allowing users to dynamically adjust parameters and instantly evaluate customized spatial clusters.
