Behavlet Analytics for Player Profiling and Churn Prediction

Abstract

Players exhibit varying behaviour from each other when playing games. Indeed, a player’s own behaviour will change as they learn to play and changing behaviour may also be indicative of when they are prematurely about to quit playing permanently (known as ‘churn’). There can be many reasons for player churn including finding a game too easy, too hard, or just not understanding what they must do. The accurate prediction of player churn is important as it allows a publisher or developer to understand and intervene to improve retention and thus increase revenue. Profiling player behaviours through their actions in-game can facilitate personalization, by adapting gameplay for different types of player to enhance player enjoyment and reduce churn rate. Behavlets are data-features that encode short activity sequences (‘atoms’ of play), which represent an aspect of playing style or player personality traits, e.g. aggressiveness or cautiousness tendencies. Previously we have shown how Behavlets can be used to model variation between players. In this paper, we focus on Behavlet sequences and how process mining and entropy-based analysis can profile evolving behaviour, predict player churn, and adapt play to potentially increase enjoyment.