Skill-based simulation model for optimizing strategy in golf

Not only skill but also determination of strategy (choosing a club and a target) is important to improve golf scores. This study aims to optimize a strategy to help golfer´s improve their scores. To optimize a strategy, first, we develop a simulation model that determines the drop position of a golf ball probabilistically. The size of a gap between a target and a position at which the ball actually falls changes depending on the skill parameters. Next, expected values of the score are calculated at every spot among the golf course by the Q-learning [9] using the simulation model. Finally, we check the capability to acquire a strategy through simulation that directs action to a golfer agent using optimized Q-tables. Results of the optimization show that when the shot distance of each club is the same but the shot accuracy is not, a club that has a shorter shot distance is more suitable for a golfer with low shot accuracy than a golfer with high shot accuracy. The strategy is optimized in consideration of course layout and golfer´s skills.