Building a Behavioral Profile and Assessing the Skill of Video Game Players

Abstract

Electronic sports (eSports) and video gaming is a rapidly evolving industry attracting billions of players and thousands professional eSports athletes worldwide. With the growing interest to this domain there is a number of challenges for players on how to progress from the amateur gaming level into the professional one and how one can characterize a professional eSports athlete and his skills. In this article, we present an answer to the essential eSports question – what underlies the players’ proficiency and defines a player as an individual distinguishing one from another. We address this problem by the investigation of the behavioral telemetry of players’ actions directly from the computer keyboard and mouse. This kind of data provides pristine characteristics of play style and represents the player’s mechanics in detail. In particular, we investigate: i) principle sequences of key controls pressed that the players use during the game, ii) sequences distinguishing the players among three groups: professionals, hardcore amateurs and casual players, iii) individual characteristics of players in different groups. To address these problems, we apply both unsupervised methods including Principal Component Analysis and supervised ensembles of Decision Trees with Recursive Feature Elimination (RFE) as feature selection technique. Our results show that the main principle factor of players’ key control mechanics reflects mostly “whether a player is a professional” rather than a general gaming skill.