Online play, online connection: A Longitudinal Social Network Analysis of BZFlag

John Paolillo and Daniel Kutz

Abstract

In this paper we investigate individual and collective activity patterns in BZFlag, an open-source, online, multi- player, 3D tank-battle game, with tens of thousands of active users worldwide. We present analysis of nearly two full years' worth of this data, sampled at five minute intervals. The results of this analysis reveal a highly-skewed distribution of activity among players, where the vast majority of players have relatively limited activity on the serv- ers. A distinct minority of players log a large number of playing hours, and exhibits a highly-structured pattern of interaction. In the aggregate picture, there is highly robust periodic activity on BZFlag at several levels. On the local scale, daily patterns predominate, with significant half-day, 2-hour and shorter periodic patterns. On the global scale, equally robust long-term periodic fluctuation is observable. Yearly patterns, clearly pertaining to the ebb and flow of the North American school year, are strongly evident. The results of this study underscore the im- portance of the time dimension in understanding patterns of social activity in online games, and in understanding social relationships more generally.