Beliefs that Entertain
|Speaker:||Joshua Tasoff, Claremont Graduate University|
|Date:||Friday 12 November 2021|
Abstract: Leisure is an important good, comprising a large fraction of people’s time-use, but it is relatively under-explored. We use a massive video game data set with 2.8 million matches from League of Legends, a two-team competitive game. For each match, we construct a belief path of which team will win by estimating a minute-by-minute predictive model. Several economic theories predict that people have preferences for various aspects of the belief path. We show that future player engagement is indeed affected by the belief path in ways that are consistent with some theories and less so with others. These results naturally lead to the question, how can one optimize the game such that it produces a distribution of game paths that maximizes some performance metric? We deconstruct the game into a data-generating process and use an evolutionary algorithm to find a locally optimal version of the game.