Abstract
There is a lot of previous research on strategies in role-playing games(RPG). The strategies can be used in a beginner support function that recommends appropriate actions to beginners. By comparing the strategy and subjects’ playing histories, it is expected that the game setting that creates the enjoyment of RPG will be discovered. The sophistication of the enemies’ strategies is expected to increase the satisfaction of improved players. But in the previous research, enemies have only simple strategies. In this research, a strategy in RPG with adaptive actions of enemies is studied. In the proposed method, actions are selected based on the max-min criterion. The enemy minimizes the player’s expected total rewards. The player maximizes the minimum expected total rewards. Markov decision processes is used in modeling. In the proposed method, dynamic programming is used. The effectiveness of the proposed method is shown by some computational examples. Since the RPG of the proposed method is more difficult than the RPG of a comparison target with an enemy’s simple strategy, the expected total rewards of the proposed method is smaller than that of the comparison target.
