Online Gamers Classification Using K-means

In order to achieve flow and increase player retention, it is important that games difficulty matches player skills. Being able to evaluate how people play a game is a crucial component for detecting gamers strategies in video-games. One of the main problems in player strategy detection is whether attributes selected to define strategies correctly detect the actions of the player. In this paper, we will study a Real Time Strategy (RTS) game. In RTS the participants make use of units and structures to secure areas of a map and/or destroy the opponents resources. We will extract real-time information about the players strategies at several gameplays through a Web Platform. After gathering enough information, the model will be evaluated in terms of unsupervised learning (concretely, K-Means). Finally, we will study the similitude between several gameplays where players use different strategies.