Optimal threat strategies of bimatrix games

A bi-matrix threat game is defined as a triple (A,B,S) whereA andB arem×n payoff matrices, andS is a closed convex subset of the plane, with (a _ij,B _ij) εS for eachi,j. Given (threat) mixed strategiesx andy,Nash's model suggests that the eventual outcome will be that point (u, v) εS which maximizes the product (u ?xAy ^t) (v ?xBy ^t) subject tou ≥xAy ^t,v ≥xBy ^t. Optimality of the threat strategies is then defined in the obvious way. A constructive proof of existence of optimal threat strategies is given; in particular, it is shown that they are optimal strategies for the matrix gameA-kB, wherek is to be determined. In this paper,k is approximated by aNewton-Raphson technique. Two examples are solved in detail.