This document discusses game theory concepts including perfect information, behavioral strategies, Nash equilibrium, and best response correspondence. It defines perfect information as each information set containing a single decision node. It also states that a Nash equilibrium always exists for finite games and provides proofs for the existence of pure strategy Nash equilibrium in games of perfect information and mixed strategy Nash equilibrium in all finite games using Kakutani's fixed point theorem and the properties of the best response correspondence.
This document discusses game theory concepts including perfect information, behavioral strategies, Nash equilibrium, and best response correspondence. It defines perfect information as each information set containing a single decision node. It also states that a Nash equilibrium always exists for finite games and provides proofs for the existence of pure strategy Nash equilibrium in games of perfect information and mixed strategy Nash equilibrium in all finite games using Kakutani's fixed point theorem and the properties of the best response correspondence.
This document discusses game theory concepts including perfect information, behavioral strategies, Nash equilibrium, and best response correspondence. It defines perfect information as each information set containing a single decision node. It also states that a Nash equilibrium always exists for finite games and provides proofs for the existence of pure strategy Nash equilibrium in games of perfect information and mixed strategy Nash equilibrium in all finite games using Kakutani's fixed point theorem and the properties of the best response correspondence.
• Perfect information: each information set contains a single decision node.
• A strategy must be a complete contingent plan that says what a player will do at each of her information sets if she is called on to play there. • (Kuhn) For games of perfect recall, there is a mixed strategy with the same distribution over outcomes as any behavioral strategy, and vice versa. – The absent-minded driver game exemplifies this, since it is not a game of perfect recall. • A Nash equilibrium always exists for finite games. – (Zermelo) In a finite game of perfect information, there exists a pure strategy NE. ∗ Proof works by induction: replace the last node with the player’s maximizing move at that node. – (Nash) Every finite game has a (mixed strategy) NE. ∗ Proof uses Kakutani’s fixed-point theorem. ∗ Define i’s best-response correspondence: ψi (p−i ) = {pi ∈ ∆Si |gi (pi , p−i ) ≥ gi (p0i , p−i ) for all p0i } · ψi is upper hemicontinuous: continuity of the payoff function ensures that for any convergent sequence of strategies in ψi , the limit will also be in ψi · ψi is convex-valued: this follows from the quasiconcavity of the payoff function. and nonempty.