Batting Order Optimization by Genetic Algorithm

Baseball has been widely studied in various ways, including math and statistics. In a baseball game, an optimized batting order helps the team achieves greater number of runs in a season. This paper introduces a method that combines a genetic algorithm with a statistical simulation to identify a non-optimal batting order. The biggest issue is how we evaluate a batting order. There are past works using dynamic programming to calculate the plate appearance and using Markov Chain to evaluate a batting order. These two algorithms summarize all past data to deliver an optimal batting order. The GA described here applies an evaluation function using a baseball game simulation. Thus the GA is more like a helping tool that can be incorporated into the decision making process rather than a deterministic tool. The simulation defines the baseball game as a set of events. By using only a subset of the event set, the decision maker can pursue a customized batting order.