The outcomes of many team-based competitions are determined by the steady accumulation of minor interactions over time. Yet we lack a general understanding of their fundamental dynamics and how those dynamics are shaped by a competition's structure.
In a recent lunchtime seminar at SFI, External Professor Aaron Clauset introduced a novel generative model that can directly quantify a competition's scoring tempo and balance. He then investigates the link between competition structure and dynamics by applying this framework to a novel data set of nearly 1 billion competitive interactions across more than 10 million diversely structured team competitions from a popular online game.
Clauset, an associate professor of computer science at UC-Boulder, and UC-Boulder collaborator Sears Merritt have posted a paper about this work on arXiv.org.
Despite wide variations in competition geography, rules, resources, and team skill, Clauset finds a common three-phase pattern in the tempo of events, a pattern also observed in some professional sports, and that competition winners are often predictable after only a few moments of gameplay.
Furthermore, tempo and balance dynamics are highly predictable from structural features alone.
Counterintuitively, the most balanced outcomes arise from specific geographic heterogeneities, not from teams of equal skill competing in homogeneous environments. These results shed new light on the principles of balanced competition, and illustrate the rich potential of online game data for investigating social dynamics and competition.
Watch the SFI lunchtime seminar video (59 minutes)