Poor convexity and Nash equilibria in games (Q2248912)

This paper is concerned with two-player normal-form games on the unit square. Continuous two-person games on the unit square such that one player's payoff function is concave in the player's own strategy possess a Nash equilibrium in which this player chooses a pure strategy and the other player mixes between two pure strategies. This was shown in [\textit{T. Parthasarathy} and \textit{T. E. S. Raghavan}, SIAM J. Control 13, 977--980 (1975; Zbl 0344.90046)]. On the other hand, if player one's payoff function is convex in the player's own strategy and the players' payoff functions are continuous in player two's strategy when player one chooses either 0 or 1, then there is a Nash equilibrium in which both players employ a mixture of two pure strategies. This was shown in [\textit{T. Radzik}, Int. J. Game Theory 21, No. 4, 429--437 (1993; Zbl 0799.90129)]. This paper extends these two results in a number of directions. First, the author proves refinements of the two results by weakening the concavity in own strategies to a condition termed \textit{pairwise poor concavity} (which is stronger than quasiconcavity in own strategies). Second, it is shown that when both payoff functions are continuous and satisfy the said weakening of concavity in own strategies, then for each mixed-strategy Nash equilibrium there is a corresponding ``simple'' equilibrium profile -- one in which the players employ a mixture of two pure strategies -- leading to the same distribution of payoffs.