Equilibrium Tracing in Bimatrix Games

We analyze the relations of the van den Elzen-Talman al- gorithm, the Lemke-Howson algorithm and the global Newton method introduced by Govindan and Wilson. It is known that the global Newton method encompasses the Lemke-Howson algorithm; we prove that it also comprises the van den Elzen-Talman algorithm, and more generally, the linear tracing procedure, as a special case. This will lead us to a dis- cussion of traceability of equilibria of index +1. We answer negatively the open question of whether, generically, the van den Elzen-Talman algorithm is ∞exible enough to trace all equilibria of index +1. In this paper we investigate several algorithms for the computation of Nash equi- libria in bimatrix games. The Lemke-Howson and the van den Elzen-Talman algorithms are complementary pivoting methods; both have been studied exten- sively. The difierence between the two methods is that while the Lemke-Howson method only allows for a restricted (flnite) set of paths, the van den Elzen- Talman algorithm can start at any mixed strategy pair, called prior, and hence generates inflnitely many paths. This implies that the van den Elzen-Talman algorithm is more ∞exible than the Lemke-Howson method. An even more ver- satile algorithm is the global Newton method (1); it works for the more general case of flnite normal form games. We investigate the relations between those three algorithms: We show that the Lemke-Howson and van den Elzen-Talman algorithms difier substantially. However, both can be understood as special cases of the global Newton method. For the van den Elzen-Talman algorithm, this is a new result, which can be generalized to the statement that for N-player games, the global Newton method implements the linear tracing procedure introduced by Harsanyi (3). As a special case of the global Newton method, the van den Elzen-Talman algorithm can generically flnd only equilibria of index +1. This leads us to the ? Supported b y the EPSRC and the LSE Research Studentship Scheme. The author would like to thank Bernhard von Stengel for helpful comments and stimulating discussions.

[1]  B. Stengel,et al.  COMPUTING EQUILIBRIA FOR TWO-PERSON GAMES , 1996 .

[2]  S. Smale Convergent process of price adjust-ment and global newton methods , 1976 .

[3]  B. Moldovanu,et al.  Understanding Strategic Interaction: Essays in Honor of Reinhard Selten , 2011 .

[4]  C. E. Lemke,et al.  Equilibrium Points of Bimatrix Games , 1964 .

[5]  J. Mertens,et al.  ON THE STRATEGIC STABILITY OF EQUILIBRIA , 1986 .

[6]  Roger B. Myerson Sustainable Equilibria in Culturally Familiar Games , 1997 .

[7]  Robert Wilson,et al.  A global Newton method to compute Nash equilibria , 2003, J. Econ. Theory.

[8]  Dolf Talman,et al.  An Algorithmic Approach toward the Tracing Procedure for Bi-matrix Games , 1999 .

[9]  Martin Shubik,et al.  A theorem on the number of Nash equilibria in a bimatrix game , 1997, Int. J. Game Theory.

[10]  L. Shapley A note on the Lemke-Howson algorithm , 1974 .

[11]  P. Jean-Jacques Herings,et al.  Homotopy methods to compute equilibria in game theory , 2006, Equilibrium Computation.

[12]  Bernhard von Stengel,et al.  Computing Normal Form Perfect Equilibria for Extensive Two-Person Games , 2002 .

[13]  Dolf Talman,et al.  A procedure for finding Nash equilibria in bi-matrix games , 1991, ZOR Methods Model. Oper. Res..

[14]  J. Harsanyi The tracing procedure: A Bayesian approach to defining a solution forn-person noncooperative games , 1975 .

[15]  John C. Harsanyi,et al.  Общая теория выбора равновесия в играх / A General Theory of Equilibrium Selection in Games , 1989 .