Automatic FPS Balance Adjustment Test Method Considering Play Style

Abstract

Existing research on automatic difficulty balancing tests for FPS using AI had problems with the accuracy of the balancing test because it could not properly capture difficulty variations due to differences in play styles. In the proposed method, multiple AI agents with specific play styles created by combining reinforcement learning and rule-based agents are introduced. Through verification, we show that it is capable of detecting difficulty changes between different play styles as well as human players.