Modeling of fighting game players

Abstract

A dynamic 7-dimensional skill-capturing game player model and accompanying novel real-time assessable metrics are introduced and validated by AI agents as well as human players. The model’s dimensions quantify player’s: 1) cognitive skills – distance estimate (DE), muscle memory (MM), reaction time (RT), space control (SC) and timing precision (TP); and 2) playing style – aggressiveness (AG) and decision making (DM). The games with AI agents indicate that methods proposed for metrics measurement are highly accurate – the anticipated outcome was achieved in 99.3 % of cases. Experiments with 16 human participants confirmed a significant correspondence between the methods’ implementation and human perception of the metrics for AG, DM and SC. Moreover, the dimensions were used to estimate the challenge factor of our in-house developed fighting game. The estimated result indicate that DM, AG, RT and SC have the greatest effect on game’s challenge; together constituting 70%. The final results of this study show that this model is very promising for applications requiring extensive behavioral and skill-capturing player characterization.