Identifying a Pitch Type that Minimizes Probability of Conceding a Score Using Monte-Carlo Simulation

Although each sport has different characteristics, the main goal of a sports game is to win by scoring at least one point more than the opponent, not about creating a large score difference. With this in mind, this study analyzes the optimal pitch type that minimizes the probability of conceding a score rather than minimizing runs in a baseball game. We represent situations in a baseball game as the Markov Decision Processes (MDP) model by calculating state transition probability according to pitch type selection based on the data recorded in the US Major League Baseball between 2021 and 2022. A formulated MDP model is used to run a Monte-Carlo simulation to identify a pitch type that minimizes the probability of conceding a score and a pitch type that minimizes total runs until the end of the inning for all states. Simulation results show that there exist differences between identified pitch types and mismatch increases in states of two outs or no runners on base. We also found that selecting pitch type to minimize the probability of conceding a score can be useful when starting an inning in a tied game.