Reinforcement Learning Algorithm to Optimize Players' Tactical Decisions and Round Planning in Tennis Matches

Abstract

This paper explores the application of the Ranking Genetic Optimization Reinforcement Learning (RGORL) algorithm to optimize players' tactical decisions and round planning in tennis matches. Leveraging evolutionary principles and reinforcement learning techniques, RGORL offers a data-driven framework for enhancing on-court performance. Extensive simulations demonstrate the algorithm's effectiveness in improving match outcomes, points won percentages, and games won percentages. Results illustrate a steady improvement in fitness scores over successive generations, indicating RGORL's ability to evolve and refine strategies over time. Analysis of tactical decisions reveals the superiority of strategies such as the "Net Approach" in terms of win rates, points won percentages, and games won percentages. Through extensive simulations, RGORL demonstrates a notable improvement in match outcomes, with a maximum increase of 13% in win rates. Analysis of tactical decisions reveals significant enhancements in points won percentages, with improvements of up to 34% observed across various strategies, notably the "Net Approach." Furthermore, the algorithm achieves substantial gains in games won percentages, with increases of up to 25% recorded.