Optimizing Sports Strategies: Achieving Convergence using SARSA and Q-Learning for Dynamic Decision-Making

Aim: The aim of this analysis is to use reinforcement learning (RL) algorithms to optimize sports strategy through improved decision accuracy. This study contrasts the effectiveness of SARSA and Q-learning with Monte Carlo methods in providing convergence and real-time strategic decision-making in dynamic sporting environments. Materials and Methods: Two cohorts were assessed in this study. Group 1 employed Monte Carlo methods to optimize decision-making techniques with episodic learning by carrying out 25 simulation-based sport scenarios. Group 2 utilized SARSA and Q-learning that incorporated temporal difference learning to adaptively optimize strategies in the same 25 scenarios. The ultimate goal was comparing and validating these methods against optimal decisionmaking accuracy. The G Power value was set at 80%, threshold at 0.05% and confidence interval at 95%. Results: SARSA and Q-learning reinforcement learning algorithms performed higher accuracy and convergence rates in the optimization of sports strategies. Group 2 achieved an accuracy of 91%, which was greater than Group 1, with Monte Carlo methods achieving an accuracy of 79%. The ability of SARSA and Q-learning to update action values in real time resulted in faster convergence and improved decision-making. Significance value is 0.05 Conclusion: The findings reveal that SARSA and Q-learning significantly outperform Monte Carlo methods in dynamic sports decisionmaking. Their real-time adaptability, convergence rate improvements, and strategic accuracy give them superior methods for sports strategy optimization. These results suggest that reinforcement learning can contribute significantly to developing AI-based decision-support systems for sports coaching and analytics.