Minimum memory for equivalence between \textit{ex ante} optimality and time-consistency (Q700118): Difference between revisions

The paper deals with the problem of the strategies timing in the extensive games. There exist two basic approaches to this phenomenon: the ex-ante optimality view where the player fixes his strategy before the actual play of the game, and the optimal time-consistency view where the player can change his strategy after each of his moves. It is known that both approaches are equivalent for the games with perfect recall and that they lead to different behaviour in some types of the inperfect recall. The main problem solved in the referred paper is the specification of an exact rate between the perfect recall and ``absent-mindedness'' under which the equivalence between both views vanishes. Such type of recall is formulated as a condition of so called A-loss recall which is still sufficient for the equivalence, and it is necessary for it in the sense that whenever it is violated, there is some pay-off assignment for which the equivalence beaks down. Moreover, the problem of the existence of a time-consistent strategy is considered, and the relation between that existence and the condition of A-loss recall is formulated.