Group evolution patterns inworld of warcraft

We study the temporal evolution of different types of guilds in the massively multiplayer online role playing gameWORLD OF WARCRAFT R# . Given a large corpus of observations of online activities of players, we apply convex-hull non-negative matrix factorization to cluster our data of about 1,400,000 guilds into well interpretable prototypes. Given these prototypes, we analyze guild formation patterns on American and European servers. We find growth patterns according to power laws that result in similar structures in both cases.