Phys. Rev. E 91, 062815 (2015)

Many a statistical physicist looking for stochastic processes to study will have considered competitive team sports and wondered about the dynamics at play, and perhaps even thought about how they might best be modelled and understood. Appreciating their complexity is one thing, but making meaningful insights that might translate into quantitative predictive capability is altogether more challenging.

Aaron Clauset and colleagues have focused on a simple, yet decisive, characteristic of sporting contests, namely the times in a game when the lead changes. Using random walk theory, they showed that the number of lead changes in any individual game follows a Gaussian distribution. More surprisingly, they also show that the probability that the last lead will change and the time when the largest lead size occurs follow the same bimodal distribution that diverges at the start and the end of the game. Finally, they evaluated the probability that any given lead is 'safe' and maintained until the game is over.

Most importantly for sports fans, the authors demonstrated a good agreement with an exhaustive body of empirical data, particularly in the case of professional basketball, where the number of scoring events in each game is very high. Sport isn't random; it's a random walk.