Quantal Response Equilibrium

The quantal response equilibrium (QRE) is a solution concept in game theory that relaxes the assumption of perfect best-response by allowing players to make errors. Introduced by Richard McKelvey and Thomas Palfrey in 1995, QRE models players as probabilistically choosing better strategies more often than worse ones, with the probability determined by a logistic (logit) function. The precision parameter λ controls how close behavior is to pure best-response: as λ → ∞, QRE converges to Nash equilibrium; as λ → 0, play becomes uniformly random.

QRE is not merely a model of noise. It is a theory of how equilibrium emerges from boundedly rational behavior. Unlike Nash equilibrium, which assumes players exactly best-respond to exact beliefs, QRE assumes players best-respond to noisy beliefs about noisy behavior — and the equilibrium is the fixed point of this mutual perturbation. The framework has been extraordinarily successful in fitting experimental data from coordination games, voting games, and auctions, often predicting out-of-sample behavior that Nash equilibrium cannot.

The deeper question is whether QRE captures something real about human cognition or merely provides a flexible curve-fitting tool. Critics note that the logit form is arbitrary — other error structures produce different equilibria — and that the precision parameter is typically estimated post-hoc. But defenders argue that the logit form is the maximum-entropy distribution given expected-payoff information, making it the least arbitrary choice possible.