Notes on a fitness solution for Nash equilibria in large games

In game theory, every equilibrium can be characterized by a binary relation on the game strategies. This relation - called generative relation - induces an appropriate solution concept. Game equilibrium is described as the set of non dominated strategies with respect to the generative relation. Evolutionary algorithms do not perform well for large games (hundred of players). This may be either because their poor scalability or because of some properties of the generative relation. Such possible properties are investigated in this paper with relation to the Pareto dominance used in multiobjective optimization. Some numerical results on two Cournot models illustrate the situation.