Pushmeet Kohli, Yoram Bachrach, David Stillwell, Michael Kearns, Ralf Herbrich, and Thore Graepel
We study how social relations between people affect the way they play the famous resource allocation game called Colonel Blotto. We report the deployment of a Facebook application called “Project Waterloo” which allows users to invite both friends and strangers to play Colonel Blotto against them. Most previous empirical studies of Blotto have been performed in a laboratory environment and have typically employed monetary incentives to attract human subjects to play games. In contrast, our framework relies on reputation and entertainment incentives to attract players. Deploying the game on a social network allows us to capture the social relations between players and analyze their impact on the used strategies. Following Arad and Rubinstein we examine player strategies and contrast them with game theoretic predictions. We then investigate how strategies are affected by social relations. Our analysis reveals that knowledge of the opponent affects the strategies chosen by players and how well they perform in the game. We show that players with few Facebook friends tend to play more games and have a higher probability of winning, that players responding to a challenge in the game have a higher probability of winning than those initiating the game, and that the initiators of a game have a higher probability of defeating their friends than strangers.
In ACM Web Sciences 2012
Publisher ACM Conference on Web Sciences