Game Solutions And The Normal Form

Abstract

Among the solution concepts proposed for noncooperative games, some can be defined on the normal formof the game whereas others can be defined only on the extensive form(or on the agent normal form, which is a game form intermediate between the normal and the extensive forms). For example, Nash equilibria and some suggested refinements of Nash equilibria can be recognized by looking only at the normal form: this property I will call normal-form dependence, Other refinements of Nash equilibria lack this property. Kohlberg and Mertens have argued that what I have called normal-form dependence should be regarded as a basic requirement for allsolution concepts for noncooperative games. I will try to show on some examples that this would not be a reasonable requirement because it would often conflictwith backward-induction rationality, which Kohlberg and Mertens themselves recognized as being an essential aspect of game-theoretic rationality.