Date of Defense

4-18-2018

Date of Graduation

4-2018

Department

Mathematics

First Advisor

Steven Ziebarth

Second Advisor

Joseph McKean

Third Advisor

Tom Tango

Abstract

In 2017, Statcast, a company that collects and analyzes major league baseball statistics, introduced a new metric called hit probability. This metric gives the likelihood that a ball will land for a hit based on the launch angle and exit velocity of the batted ball. One limitation of the Statcast model is with the home run breakdown of hit probability; balls hit to any part of the field are given the same home run probability. This does not take into account the difference in outfield lengths at different angles. With this in mind, I have created a model based on hit distance and spray direction to model home run probability. The model is based on logistic regression, where a home run is considered a success and anything else is considered a failure. These probabilities provide instant context to batted balls, and are free from the influence of which park the ball was hit, and which fielders it was hit against. Examples from the 2017 season support the need for spray direction to be used in a model when discussing home run probability. Overall, the model shows good success as a tool for modeling home run probability and may help baseball fans of many kinds to better understand the game as it unfolds in front of them.

Access Setting

Honors Thesis-Open Access

Additional Files
PRESENTATION.pdf (9827 kB)
Defense Presentation
Download

Share

COinS