Online "Guilds" Mirror Real Life Gangs
j-beda writes "In June 2009, Dr. Neil Johnson published a paper titled 'Human group formation in online guilds and offline gangs driven by a common team dynamic' in Physical Review E that found the way in which WoW 'guilds' form can be described by a mathematical model that can also be applied to an unrelated group of people: street gangs in Los Angeles. Since 'Any group that satisfies these fairly autonomous, competitive criteria would also (fit the model),' said Dr. Johnson, the findings are of interest to those combating international as well as local terrorist cells."
You might be interested: Robber's Cave Experiment
Another (not a scientific) study: The Third Wave
They measure cumulative size distribution (how many groups of size >= N) and churn (how many people leave the group for another one in a given period).
They are able to come up with a simple mathematical model for the behaviour of players (essentially: recruit people with diverse attributes/skills) that reproduces the observed data extremely well. And they also show that the alternative 'kinship' model (recruit people with similar attributes/skills) fails to reproduce the observed data.
I would say that their model does quite a good job at modeling some rather nontrivial data.
http://lanl.arxiv.org/abs/0812.2299 is a better (free) link to the preprint.