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 videogames. 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, KMeans). Finally, we will study the similitude between several gameplays where players use different strategies.