Blog · Collection

Game theory

Strategic reasoning, from Nash equilibrium and mixed strategies to backward induction, bargaining, repeated games, signaling, auctions and Bayesian games.

Game theory studies decisions where the best choice depends on what others choose. This overview moves from simultaneous games to dynamic ones and finally to games with private information.

Simultaneous-move games

A normal-form game lists the players, their strategies and their payoffs. Eliminating dominated strategies already solves some games, like the prisoner’s dilemma, where individually rational play leaves both players worse off. A Nash equilibrium is a combination of strategies in which each player’s strategy is a best response to the others’. Classic applications include political candidates converging on the median voter and firms choosing quantities in a Cournot duopoly. Games can have several equilibria, some better for everyone than others, which raises the question of how players coordinate.

Mixed strategies and evolution

Rock–paper–scissors has no equilibrium in pure strategies. In its mixed equilibrium, each player randomizes so as to make the others indifferent among their choices; penalty kicks and tennis serves follow the same logic. Evolutionary game theory reinterprets strategies as inherited traits and asks which ones resist invasion by mutants, which leads to evolutionarily stable strategies.

Dynamic games

In extensive-form games, a strategy is a complete contingent plan. Backward induction solves games of perfect information, and Zermelo’s theorem guarantees that games like chess are determined. Subgame perfection refines Nash equilibrium by ruling out threats that a player would not actually carry out. In alternating-offer bargaining, patience is bargaining power, and wars of attrition show how costly delay can persist in equilibrium.

Repeated games

A one-shot prisoner’s dilemma ends in defection, and by backward induction so does any fixed, finite number of repetitions. With an infinite horizon, or a random ending, and patient enough players, cooperation can be sustained by strategies such as grim trigger and tit-for-tat, which punish deviations in later rounds.

Private information

When players know things that others don’t, their actions carry information. In signaling games, a better-informed player may take a costly action to reveal its type. Auctions bring the winner’s curse: in a common-value auction, winning means your estimate was probably the most optimistic one. Bayesian games formalize all of this, and a Bayesian Nash equilibrium specifies a best response for every possible type of every player.

← All collections