[3H1-GS-3] Data mining: Applied data mining (1)

[3H1-GS-3-05] A Prediction Model of Earned Runs based on Latent Class Markov Chain for Starters of Professional Baseball Pitchers

〇Ryosuke Uehara¹, Takuma Matsumoto¹, Kenta Mikawa², Masayuki Goto¹ (1. Waseda University, 2. Shonan Institute of Technology)

Keywords:baseball, starting pitcher, expected runs, latent class model, markov chain model

In recent years, quantitative analysis for baseball is performed using a large amount of accumulated data for various purposes, such as a novel defensive shift and evaluation of players. This paper proposes a predictive model of the expected runs for each inning of starting pitchers. At that time, we apply the latent class model and group the combinations of the pitcher and the batter by a small number of latent variables, and the \lq\lq batting average\rq\rq is calculated for each latent class. Consequently we construct a model to calculate expected runs considering the difference of "batting average" between pitcher and batter matches. To verify the effectiveness of the proposed method, we conduct experiments by using actual Japanese professional baseball data.