Regret-based continuous-time dynamics. (Q1416502)

The analysis of the dynamics of the play of a given game when players act in accordance with certain strategies has acquired special relevance in the recent years. This paper provides a continuous-time analysis of the dynamics associated with regret-based strategies. More specifically, the authors show that the main findings of the papers [J. Econ. Theory 98, 26--54 (2001; Zbl 0994.91007)] and [Econometrica 68, 1127--1150 (2000; Zbl 1020.91003)] for the discrete-time setting also carry out with continuous time. Indeed, as the authors emphasize, this new setting allows for a ``streamlined treatment of the dynamics''. The main results included in this paper are summarized below. i) Under regret-based strategies, the time-average joint play converges to the Hannan set (analog of the discrete-time result). Moreover, for either two-person zero-sum games or two-person potential games the convergence is to the set of Nash equilibria. ii) Under conditional regret-based strategies, the time-average joint play converges to the set of correlated equilibria (essentially analogous to the distrete-time result).