Interpretable Real-Time Win Prediction Model Based on Health for Fighting Games

Abstract

In this paper, we propose an interpretable real-time win prediction model in fighting games. Previous studies about match of the win prediction in e-Sports have focused on MOBA, primarily using a method that defines the progress of the game based on time. However, this method may not always be optimal for fighting games due to variability in round durations. Therefore, we propose a model that defines the progress of the game based on health instead of time, and show its effectiveness.