Batted ball spray charts: a system to determine infield shifting

Abstract

To gain a competitive advantage, the Tampa Bay Rays Baseball Club uses data to inform a range of decisions. This project focuses on forecasting batted ball trajectories across the infield using historical data, by outputting a likelihood probability for nine different infield zones. The paper details the development of a model that can predict where a baseball is likely to land based on a number of variables, most importantly the particular batter and pitcher. Using historical batted ball data, the back-end algorithm was developed using mixed effect models. Due to the proprietary nature of the algorithms, most of the paper will focus on the front-end design and human factors testing of the system. The fully functional user interface provides batted ball spray charts of the infield for batter pitcher combinations. Rays personnel will quickly be able to view the predicted probability for each section of the infield. Based on a set of requirements developed by interviewing Rays personnel, the system was tested with multiple groups of users to determine information extraction rates and ease of use. The test results for different versions of the decision tool helped drive the final application. The Rays intend to use this system for the 2015 MLB regular season to effectively position their infielders to convert more batted balls into outs and ultimately, win more baseball games.