lecture5 notes

# Dene the set of best response correspondences as b s

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: strategy existence in inﬁnite games. Theorem (Debreu, Glicksberg, Fan) Consider a strategic form game �I , (Si )i ∈I , (ui )i ∈I � such that for each i ∈ I 1 Si is compact and convex; 2 ui (si , s−i ) is continuous in s−i ; 3 ui (si , s−i ) is continuous and concave in si [in fact quasi-concavity suﬃces]. Then a pure strategy Nash equilibrium exists. 20 Game Theory: Lecture 5 Existence Results Deﬁnitions Suppose S is a convex set. Then a function f : S → R is concave if for any x , y ∈ S and any λ ∈ [0, 1], we have f ( λx + (1 − λ )y ) ≥ λf (x ) + (1 − λ )f (y ) . concave function not a concave function 21 Game Theory: Lecture 5 Existence Results Proof Now deﬁne the best response correspondence for player i , Bi : S−i � Si , � � Bi (s−i ) = si� ∈ Si | ui (si� , s−i ) ≥ ui (si , s−i ) for all si ∈ Si . Thus restriction to pure strategies. Deﬁne the set of best response correspondences as B (s ) = [Bi (s−i )]i ∈I . and B : S � S. 22 Game Theory: Lecture 5 Existence Results Proof (continued) We will again apply Kakutani’s theorem to the best response correspondence B : S � S by showing that B (s ) satisﬁes the conditions of Kakutani’s theorem. 1 S is compact, convex, and non-empty. By deﬁnition S= ∏ Si i ∈I since each Si is compact [convex, nonempty] and ﬁnite product of compact [convex, nonempty] sets is compact [convex, nonempty]. 2 B (s ) is non-empty. By deﬁnition, Bi (s−i ) = arg max ui (s , s−i ) s ∈ Si where Si is non-empty and compact, and ui is continuous in s by assumption. Then by Weirstrass’s theorem B (s ) is non-empty. 23 Game Theory: Lecture 5 Existence Results Proof (continued) 3. B (s ) is a convex-valued correspondence. This follows from the fact that ui (si , s−i ) is concave [or quasi-concave] in si . Suppose not, then there exists some i and some s−i ∈ S−i such that Bi (s−i ) ∈ arg maxs ∈Si ui (s , s−i ) is not convex. This implies that...
View Full Document

## This document was uploaded on 03/19/2014 for the course EECS 6.254 at MIT.

Ask a homework question - tutors are online