Sequential Analysis of Player Behavior

Abstract

Understanding how interaction unfolds over time is a key factor for understanding the dynamics aspects of player behavior. Thus far, analysis of sequential patterns of player behavior has, however, mainly focused on discovering frequently recurring patterns. However, frequency of occurrence is not always a reliable indicator of a pattern's importance as infrequent patterns can also offer valuable insights about in-game behavior. In this paper we thus propose the use of lag sequential analysis (LSA) -- which, rather than relying on frequency counts, makes use of statistical methods to determine the significance of sequential transitions -- to aid analysis of behavioral streams of players. For this purpose we apply LSA to in-game data of two well-known games. The meaningfulness of the identified sequences is verified by comparing them to documented and established strategies. In addition, results obtained through LSA are discussed in relation to results from frequency-based sequence mining. Our results suggest that LSA is a promising complement to frequency based methods for analyzing sequential behavior patterns of players.