Online adjustment of the AI's strength in a fighting game using the k-nearest neighbor algorithm and a game simulator

This paper proposes a method for online adjustment of the AI's strength in a fighting game. The adjustment aim is that of maintaining the AI's strength slightly above its opponent's level. The proposed method adopts the k-nearest neighbor algorithm for predicting the opponent's next action and uses this information together with a game simulator for determining the next action of the AI, leading to slightly winning scores against the opponent. The proposed method is evaluated on a fighting-game used at the fighting game AI competition organized by the authors since 2013. Evaluation results confirm the effectiveness of the proposed method.