Mixed strategies in strictly competitive games: A further test of the minimax hypothesis

Abstract

We report the results of two studies designed to investigate interactive behavior in two-person zero-sum games and assess the descriotive power of the minimax hypothesis. Standard statistical tests reject the particular multinomial model implied by the minimax model on the individual and group levels. On the other hand, the minimax model outperforms an equiprobable model predicting equal choice probabilities and a win-weighted model in which the probability of choosing each pure strategy is proportional to the number of opportunities to win associated with this strategy. The combined results of the two studies support strategic equivalence.