A Practical Framework For Space-Time Analysis of Baseball Data

Authors: Chih-Yuan Chen*, Assistant Professor, Department of Geography, Chinese Culture University, Hsiang-Chih Yu*, Department of Geography, Chinese Culture University , Yen-Hui Tsai, Department of Geography, Chinese Culture University
Topics: Spatial Analysis & Modeling
Keywords: Baseball, Space-Time analysis, Visualization, Stereo Photogrammetry
Session Type: Poster
Day: 4/13/2018
Start / End Time: 1:20 PM / 3:00 PM
Room: Napoleon Foyer/Common St. Corridor, Sheraton, 3rd Floor
Presentation File: No File Uploaded

Sabermetrics has brought the data revolution to baseball, and shown how advanced data and analysis can help teams get wins. In 2006, Major League Baseball (MLB) first introduced stereo photogrammetry and radar technology to capture a new type of spatiotemporal baseball data which the three-dimensional position and trajectory of players and balls can be estimated. Based on our previous research, we introduce a practical framework for space-time analysis of baseball, which can collect, store, analyze, and display this new type of spatiotemporal baseball data. Frist, we use multiple cameras to collect stereo images, then using image processing technology to estimate the player’s positions and trajectories. Traditional baseball score keeping is also adopted manually to get play-by-play information. Second, we categorize all the information into three different fields: (1) game events (e.g. strike out, homerun, and fielder’s error ); (2) pitching data (e.g. ball position, velocity, and pitch type); and (3) hitting, running, and fielding data (ball trajectory, runner's position, fielder's position). We then apply different spatial and temporal granularities to each category for analysis purposes. Then, we apply space-time analysis methods (e.g. machine learning and data mining) to explore the data using different types of “time”(e.g. innings, ball counts, or game time) like spatial heterogeneity in different baseball stadiums, and kernel density estimation for ball position. Last, the collected data and the analytical results are visualized to demonstrate the capability of this framework.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login