Learning Contextual Event Embeddings to Predict Player Performance in the MLB

Research Paper will be posted in the coming weeks. Check back soon!
Download the
Full Paper Here
Authors

Connor Heaton, Prasenjit Mitra,

Abstract

This paper leverages recent advances in deep learning (DL) to contextualize events that occur on the baseball diamond. Similar to how large language models such as the popular ChatGPT learn to understand language as a sequence of words, we train a model to understand the game of baseball as a sequence of pitches. We then use this understanding of the game to make predictions about how players will perform in the future based on their previous performances. Using only 10 games worth of pitch-by-pitch data, we can make predictions for single-game pitcher strikeouts and binary batter has-hit predictions that are competitive with three major sportsbooks in the US.