MONEYBaRL: Exploiting pitcher decision-making using Reinforcement Learning
This work addresses a domain-specific problem for baseball analysts and teams by providing a data-driven approach to evaluate pitcher strategies, though it appears incremental as it applies existing reinforcement learning methods to a new sports context.
The paper tackled the problem of modeling baseball pitchers' decision-making by framing at-bat scenarios as a Markov Decision Process to exploit their 'Baseball IQ', resulting in a computational model that analyzes pitch selection and outcomes based on historical data.
This manuscript uses machine learning techniques to exploit baseball pitchers' decision making, so-called "Baseball IQ," by modeling the at-bat information, pitch selection and counts, as a Markov Decision Process (MDP). Each state of the MDP models the pitcher's current pitch selection in a Markovian fashion, conditional on the information immediately prior to making the current pitch. This includes the count prior to the previous pitch, his ensuing pitch selection, the batter's ensuing action and the result of the pitch.