Towards Detecting Clusters of Players using Visual and Game-play Behavioral Cues

The issue of discriminating among players’ styles and associating them with player profile characteristics, demographics and specific interests and needs is of vital importance for creating content, fine tuned and optimized in such a way that user engagement and interest are maximized. This paper attempts to address the issue of clustering players’ behavior using visual features and player performance, as input parameters. Following an unsupervised scheme, in this work, we utilize data from Super Mario game recordings and explore the possibility of retrieving classes of player types along with existing correlations with certain global characteristics.