"Artificial Intelligence Can Transform Baseball"

2023-06-20

The era of traditional "Whitey ball" has passed. In today's baseball games, coaches, scouts, managers, and players themselves heavily rely on quantitative analysis, which has completely transformed the sports industry in the past decade. But what if artificial intelligence could further drive these advancements? A professor from the McKelvey School of Engineering at Washington University in St. Louis has proposed a computational game theory approach that has been proven to enhance the effectiveness of both average and below-average Major League Baseball pitchers. "We essentially used a combination of stochastic game models and deep neural network learning techniques to calculate the optimal pitching sequence in baseball hitting," said Yevgeniy Vorobeychik, an associate professor of computer science and engineering and co-author of the paper "Optimal Pitching Strategies in Computational Baseball Hitting." He recently presented this achievement at the Florida Artificial Intelligence Research Society (FLAIRS) conference. Vorobeychik and his collaborators used deep neural networks to learn how to predict the outcomes of batters during hitting. They then modeled a hitting opportunity as a stochastic game and solved this game using value iteration combined with linear programming, a mathematical modeling technique. By focusing on data from the 2015 to 2018 seasons of Major League Baseball, they identified strategies for each hitting opportunity that actually helped improve the effectiveness of both average and below-average pitchers. The main findings of this study include: - Dynamic Game: Vorobeychik and his collaborators developed a dynamic game theory model that considers the pitcher's repertoire and control ability, as well as the batter's patience or tendency to swing outside the strike zone, in order to generate the optimal pitching sequence strategy for each hitting opportunity. - Data Integration: By integrating comprehensive player and game data, including historical performance and pitch tracking information, the researchers created a framework that can generate personalized pitching strategies for individual pitchers against specific batters. - Performance Analysis: The study evaluated the effectiveness of these strategies by comparing the optimized pitching strategies with the observed pitching outcomes in the data. The results showed significant improvements in the average on-base percentage of batters, particularly for lower-ranked pitchers. "I suspect that virtually all pitchers who make it to the Major Leagues have excellent pitching skills," Vorobeychik said. "The part of what distinguishes the exceptional from the merely good pitchers is how they apply their arsenal in the game environment against specific hitters. Formalizing this situation as a game may enable less experienced or skilled pitchers to identify the optimal pitching sequence and better utilize their skills." But can this model help Major League pitchers in a game environment? "I believe it can," Vorobeychik said. "Of course, there is more work to be done. For example, we assume that each batter is independent, but that is clearly not the case. This is actually what we are working on now." Source: https://techxplore.com/news/2023-06-ai-baseball.html

RECENT AI NEWS

RECENT AI TOOLS