Abstract:
|
Over the last decade the eSports industry has expanded to a scope rivaling traditional sports with no signs of slowing down. While the sport has rapidly expanded the amount of data available to analyze the game has also expanded. However, the use of advanced statistics to rank players is not as common as it is in traditional sports. In this talk, we present a regression-based method for assessing player performance in eSports using, as an example, the game Defense of the Ancients 2 (DOTA 2). Here we introduce analytical methods commonly used in traditional sports analytics and show how they can be applied to ranking professional eSport athletes. In particular, we demonstrate how Bayesian hierarchical models can both explain as well as predict the probability of a professional team being victorious. We further show how an in game metric, the amount of gold farmed for each player, can be used as a proxy for winning allowing traditional hierarchical Gaussian regression models to be used.
|