We study a class of non-cooperative games that includes many standard oligopoly games, macro economic coordination games, network and production externality games, and others. For these games, the sets of rationalizable strategies, pure Nash equilibrium strategies, and correlated equilibrium strategies are non-empty and have identical upper and lower bounds. Also, a large class of dynamic learning processes - including both best-response dynamics and Bayesian learning - lead eventually to behavior that lies between these same bounds. General comparative static and welfare theorems are provided.