Game Theory

Game theory is the study of decisions made when the outcome depends on what others do.

12 frameworks, from 1928 to today · Microeconomics · Economics

still a working view ended · each leads to its entry below

Overview

Game theory is the study of decisions made when the outcome depends on what others do1. A firm setting a price, a country negotiating a treaty, a bidder at an auction, a driver choosing a route, and two people splitting a bill all face the same structural problem: my best choice depends on what you choose, and yours depends on mine. Game theory builds formal models of such situations and asks what rational players would do, what outcomes are stable, and how the rules of the interaction shape the result. It became one of the most widely used tools in economics and spread into political science, biology, computer science, and beyond. Its history is a sequence of attempts to answer a deceptively simple question—what does it mean to act rationally when others are acting rationally too?—and each answer opened new problems.

Precursors: Cournot, Zermelo, and Borel

The earliest formal treatment of strategic interdependence came from the economist Antoine Augustin Cournot in 1838. He modeled two firms choosing quantities of a homogeneous good, each maximizing profit given the other's output12. The equilibrium he found—where neither firm wants to change its quantity given the other's—is a special case of what would later be called Nash equilibrium, though Cournot did not frame it in general terms32. His work was largely ignored by economists for decades.

In 1913 Ernst Zermelo proved that chess is either a win for White, a win for Black, or a draw with perfect play—a result about the determinacy of finite games with perfect information. Émile Borel wrote several papers around 1921–1927 on symmetric zero-sum games and the notion of a "method of play," but he did not prove the general minimax theorem and initially doubted that a value always exists. These scattered results established that strategic situations could be formalized, but no general theory existed.

Von Neumann and Morgenstern: The Founding

John von Neumann proved the minimax theorem for finite two-person zero-sum games in 192814. The theorem states that in such a game—where one player's gain is exactly the other's loss—there is a value v such that the first player can guarantee at least v and the second can hold the first to at most v, provided players may randomize their choices1. The key idea is a mixed strategy: a probability distribution over pure actions15. Randomizing makes one's play unpredictable and prevents the opponent from exploiting a pattern. In matching pennies, for example, no pure choice is safe, but choosing heads and tails with equal probability guarantees an expected payoff of zero against any opponent.

Von Neumann and Oskar Morgenstern's Theory of Games and Economic Behavior (1944) extended this into a general framework16. They introduced the extensive form (a game tree showing the order of moves and information available at each point) and the normal form (a matrix of payoffs for each combination of strategies). They also developed cooperative game theory, including the concept of a coalition—a subset of players who can make binding agreements—and the von Neumann–Morgenstern stable set, a solution concept for cooperative games1. Their book framed game theory as a foundation for economics, though its immediate influence on economics was limited; its main early impact was in military strategy and operations research during and after the Second World War.

Nash: Equilibrium and the Bargaining Problem

John Nash's 1950 doctoral thesis, published as "Equilibrium Points in N-Person Games" (1950) and "Non-Cooperative Games" (1951), provided the central solution concept. A Nash equilibrium is a profile of strategies, one per player, such that no player can improve their payoff by unilaterally changing strategy given the others' choices13. In Cournot's duopoly, the quantity pair where neither firm wants to adjust is a Nash equilibrium13. In the prisoner's dilemma, both players defecting is the unique Nash equilibrium, even though both would be better off cooperating—a result that captures the tension between individual and collective rationality.

Nash proved that every finite game has at least one Nash equilibrium, possibly in mixed strategies13. This existence theorem made the concept universally applicable and turned game theory into a general theory of strategic interaction. Nash also contributed to cooperative bargaining theory with the Nash bargaining solution (1950), which specifies a unique outcome for two players dividing a surplus based on axioms of efficiency, symmetry, scale invariance, and independence of irrelevant alternatives. The solution is the point that maximizes the product of the players' gains over their disagreement point1.

Nash's work shifted the field's center of gravity from cooperative to non-cooperative game theory—from games where players can make binding agreements to games where they cannot. The distinction matters: cooperative theory asks what coalitions can achieve and how to divide the gains, while non-cooperative theory models the strategic choices individuals make and predicts outcomes from those choices.

Refinements: Selten, Harsanyi, and the Problem of Multiple Equilibria

Nash equilibrium has a practical difficulty: many games have multiple equilibria, and the concept does not say which one players will reach. Reinhard Selten addressed this in 1965 with the concept of subgame perfect equilibrium. In an extensive-form game, a subgame is a portion of the game tree that begins at a single node and includes all subsequent moves. A strategy profile is subgame perfect if it constitutes a Nash equilibrium in every subgame7. This rules out incredible threats—threats that a player would not carry out if the situation actually arose7. In a sequential game where an incumbent threatens to fight a new entrant, the threat is credible only if fighting is optimal at the point of decision; if it is not, subgame perfection eliminates the equilibrium in which the entrant stays out because of the threat.

John Harsanyi's 1967–1968 papers on games of incomplete information addressed a different problem: players often do not know each other's payoffs, capabilities, or information. Harsanyi introduced the Bayesian game, in which each player has a type (private information) drawn from a common prior distribution. A Bayesian Nash equilibrium is a profile of strategies, one per type of each player, such that each type's strategy is optimal given beliefs about others' types derived from the prior and the strategies they play. This framework made it possible to model auctions, bargaining with private valuations, and signaling—situations where what you know about your opponent matters as much as what they do.

Repeated Games and Reputation

The prisoner's dilemma played once has a grim prediction: both defect. But played repeatedly, cooperation can emerge. In a repeated game, the same stage game is played multiple times, and players can condition current actions on past behavior18. The folk theorem for repeated games (developed in various forms by several researchers in the 1950s–1970s, with important contributions from James Friedman in 1971) states that in infinitely repeated games with sufficiently patient players, a wide range of outcomes—including cooperation—can be sustained as equilibria. The mechanism is the threat of future punishment: a player cooperates today because defecting would trigger retaliation tomorrow.

Robert Axelrod's 1980 computer tournament popularized the tit-for-tat strategy—cooperate on the first move, then copy the opponent's previous move—as a robust strategy in repeated prisoner's dilemma. The tournament showed that simple, forgiving, and retaliatory strategies can outperform more complex ones. David Kreps, Paul Milgrom, John Roberts, and Robert Wilson (1982) showed that reputation effects can sustain cooperation even in finitely repeated games if there is a small probability that some players are irrational or committed to cooperation. The chain-store paradox—the puzzle of why an incumbent would ever fight entry in a finite sequence of markets—was resolved by showing that reputation for toughness can be rational when there is incomplete information about the incumbent's type.

Cooperative Game Theory: Shapley, Core, and Matching

While non-cooperative theory dominated economics, cooperative game theory developed its own solution concepts. Lloyd Shapley introduced the Shapley value in 1953, a unique allocation of a coalition's total payoff that satisfies axioms of efficiency, symmetry, dummy player, and additivity. The Shapley value assigns each player their average marginal contribution across all possible orderings of players. It has been applied to cost allocation, voting power, and network centrality.

The core of a cooperative game is the set of allocations that no coalition can improve upon by deviating9. It generalizes the idea of Pareto efficiency to coalitional settings. The core can be empty, which limits its applicability, but it remains a central concept in matching and market design.

David Gale and Lloyd Shapley's 1962 paper on the stable marriage problem introduced matching theory. They showed that for any set of men and women with strict preferences, there exists a stable matching—one where no pair would prefer each other to their assigned partners. The deferred acceptance algorithm produces such a matching. This work later became the foundation for market design in school choice, kidney exchange, and residency matching, for which Alvin Roth and Shapley shared the 2012 Nobel Prize.

Mechanism Design: Reverse Game Theory

Mechanism design asks how to design the rules of a game so that self-interested players produce a desired outcome. It is often called reverse game theory: instead of taking the rules as given and predicting behavior, the designer chooses the rules to achieve a goal. Leonid Hurwicz introduced the concept in the 1960s and 1970s, studying incentive compatibility—the property that players have an incentive to reveal their private information truthfully.

The revelation principle, formalized by Roger Myerson (1979) and others, states that any outcome achievable by any mechanism can be achieved by a direct mechanism in which players truthfully report their types. This simplifies the design problem enormously: the designer can restrict attention to truthful direct mechanisms. Myerson's 1981 paper on optimal auction design used this principle to derive the revenue-maximizing auction, showing that the optimal mechanism often involves setting a reserve price and allocating to the highest bidder above it.

William Vickrey's 1961 paper on counter-speculation, auctions, and competitive sealed tenders introduced the second-price auction (now called the Vickrey auction), in which the highest bidder wins but pays the second-highest bid. Bidding one's true value is a dominant strategy in this auction—a result that made it a canonical example of incentive compatibility. Vickrey shared the 1996 Nobel Prize with James Mirrlees for their work on incentives under asymmetric information.

Evolutionary and Behavioral Game Theory

Evolutionary game theory emerged from biology in the 1970s. John Maynard Smith and George Price (1973) introduced the concept of an evolutionarily stable strategy (ESS)—a strategy that, if adopted by a population, cannot be invaded by a small number of mutants playing a different strategy10. The ESS is a refinement of Nash equilibrium that incorporates population dynamics. Maynard Smith's 1982 book Evolution and the Theory of Games systematized the field1110. In economics, evolutionary game theory provided a way to model learning and adaptation, and it connected to replicator dynamics, a mathematical model of how strategy frequencies change over time based on relative payoffs1.

Behavioral game theory emerged from experimental economics in the 1980s and 1990s. Researchers found systematic deviations from the predictions of standard game theory: in ultimatum games, proposers offer more than the minimum and responders reject low offers; in public goods games, contributions are higher than predicted by pure self-interest; in dictator games, many people give positive amounts. Colin Camerer's Behavioral Game Theory (2003) synthesized these findings. Models of social preferences—such as inequity aversion (Fehr and Schmidt, 1999) and reciprocity (Rabin, 1993)—modified utility functions to include concerns for fairness and the intentions of others. Level-k and cognitive hierarchy models (Stahl and Wilson, 1995; Camerer, Ho, and Chong, 2004) described players who reason iteratively about others' reasoning, with level-0 players acting naively and higher-level players best-responding to lower-level types.

Algorithmic Game Theory

The rise of the internet and large-scale computation created new questions. Algorithmic game theory studies the computational aspects of strategic interaction: how hard is it to compute equilibria? How do algorithms perform in strategic settings? How should we design mechanisms when players are computational agents? The field crystallized around the turn of the twenty-first century, with Tim Roughgarden's work on the price of anarchy—the ratio of the worst-case equilibrium outcome to the optimal outcome—quantifying the inefficiency of selfish behavior in networks and congestion games. Noam Nisan and colleagues' Algorithmic Game Theory (2007) became a standard reference. The field connects game theory to computer science, auction design for online advertising, and the study of equilibria in large games.

The Present

Game theory today is a sprawling toolkit rather than a single theory. Nash equilibrium remains the default solution concept, but it is supplemented by refinements (subgame perfection, sequential equilibrium, trembling-hand perfection) that address its multiplicity problem. Bayesian games and mechanism design are central to auction theory, contract theory, and market design. Behavioral game theory has documented robust deviations from standard predictions and built models to accommodate them. Evolutionary game theory provides a bridge to biology and to learning models. Algorithmic game theory addresses computational constraints and large-scale strategic environments.

Open questions remain. How do people actually reason in strategic situations, and can we build models that predict better than they explain? How do equilibria behave in games with many players and complex strategy spaces? How should mechanisms be designed when players are boundedly rational or when the designer's information is limited? The field that began with von Neumann's minimax theorem and Nash's existence proof has become a general language for strategic interaction—one that continues to generate new questions as it is applied to markets, politics, biology, and computation.

References

  1. Game theory (introduction, Cournot competition, As a formal discipline (early–mid 20th century), Minimax, Strategies, History, Characteristic function form, Von Neumann–Morgenstern stable set, Equilibria, Nash, Nash bargaining, Repeated, Evolutionary), Wikipedia
  2. Cournot competition (introduction, History), Wikipedia
  3. Nash equilibrium (introduction, History), Wikipedia
  4. Minimax theorem, Wikipedia
  5. Strategy (game theory) (Pure and mixed strategies), Wikipedia
  6. Theory of Games and Economic Behavior, Wikipedia
  7. Subgame perfect equilibrium, Wikipedia
  8. Repeated game, Wikipedia
  9. Core (game theory), Wikipedia
  10. Evolutionarily stable strategy (introduction, History), Wikipedia
  11. Evolution and the Theory of Games, Wikipedia

The frameworks of Game Theory

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1928-1960
  2. 1944-Present
  3. 1950-Present
  4. 1950-Present
  5. 1962-Present
  6. 1965-Present
  7. 1967-Present
  8. 1970-Present
  9. 1971-Present
  10. 1973-Present
  11. 1982-Present

    Behavioral Game Theory

    About Behavioral Game Theory
    Standard game theory begins with a stark assumption: players are perfectly rational, self-interested, and infinitely strategic. Behavioral game theory asks what happens when real people—with limited cognitive resources, social emotions, and a tendency to learn from experience—sit down to play. It does not abandon the formal tools of game theory. Instead, it uses experiments to observe how actual behavior deviates from the Nash equilibrium benchmark, and then builds models that incorporate those deviations as systematic features rather than random errors. The core problem is to explain how people think strategically, what they care about beyond their own payoffs, and how they adapt their play over time.
    The curriculum for a first course on this framework typically moves through three clusters of ideas. The first cluster covers how players form beliefs about others. Cognitive hierarchy theory and limited strategic thinking capture the finding that most people do not perform an infinite recursion of 'she thinks that I think that she thinks...' Instead, they stop after a few levels of reasoning, and the distribution of these 'levels of reasoning' is a key input to prediction. Behavioral equilibrium concepts, such as the Quantal response equilibrium (QRE), replace the assumption of perfect optimization with a probabilistic choice rule: players choose better responses more often than worse ones, but not always, and they correctly anticipate that all other players follow the same noisy logic. The second cluster addresses what people want. Social utility functions and other-regarding preferences formalize the idea that players care about fairness, reciprocity, and the distribution of outcomes, not only their own material payoff. These models explain why a responder in the Ultimatum Game routinely rejects a low offer even though standard theory says 'take anything.' The third cluster examines how players learn and communicate over time. Reinforcement learning treats strategy choice as a process of updating propensities based on past payoffs, without requiring players to understand the strategic structure of the game. Strategic teaching and sophistication look at the opposite side: players who recognize that their current actions can influence opponents' future beliefs and behavior, and who form plans accordingly. Communication and cheap talk studies how pre-play messages—even non-binding ones—can reshape expectations and outcomes. Coordination and social preferences show how shared norms or simple signals help groups select among multiple equilibria.
    The intellectual history of behavioral game theory begins in the 1980s, when experimental economists started to notice systematic anomalies in games like the Prisoner's Dilemma and the Ultimatum Game. The seminal work that launched the field as a coherent framework was 'Behavioral Game Theory: Experiments in Strategic Interaction' (2003) by Colin F. Camerer. This book synthesized two decades of experimental findings and formal models. Camerer's central insight was that the deviations from Nash equilibrium were not noise—they were patterned, predictable, and explainable by a small set of behavioral principles. Around the same time, Richard H. Thaler and others were developing 'behavioral economics' more broadly, but behavioral game theory carved out a distinct niche: it kept the mathematical structure of games but modified the assumptions about preferences and cognition. Key early models included the 'cognitive hierarchy' model of Camerer, Teck-Hua Ho, and Juin-Kuan Chong (2004), which estimated how many steps of strategic thinking players use; and the 'quantal response equilibrium' of Richard McKelvey and Thomas Palfrey (1995), which connected probabilistic choice to equilibrium reasoning. Matthew Rabin's work on 'fairness' (1993) gave game theorists a tractable way to let social preferences enter the payoff matrix. The field also drew heavily on the psychological insights of Daniel Kahneman and Amos Tversky, especially Prospect Theory (1979), though behavioral game theory applied these ideas specifically to strategic interaction rather than individual choice. Over time, the framework absorbed tools from evolutionary game theory (learning dynamics) and from experimental methodology (how to design incentives to reveal true preferences). Behavioral game theory did not replace its predecessors. It coexists with them, offering a richer toolkit for predicting behavior in laboratory and field settings, and it has become the standard approach for any applied microeconomist who wants to model how agents actually behave rather than how hyper-rational agents would behave in theory.
    The 10 concepts its tower teaches
    • Behavioral equilibrium
    • Cognitive hierarchy theory
    • Communication and cheap talk
    • Coordination and social preferences
    • Limited strategic thinking
    • Other-regarding preferences
    • Quantal response equilibrium
    • Reinforcement learning
    • Social utility functions
    • Strategic teaching and sophistication

    Fly to its tower in Nousfarer

  12. 1999-Present

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Minimax theoremCooperative Game TheoryBargaining TheoryNash equilibriumMatching and Market DesignEquilibrium Refinement TheoryBayesian gameMechanism DesignRepeated gameEvolutionary Game TheoryBehavioral Game TheoryAlgorithmic Game Theory

Beyond the field

Where Game Theory's ideas came from in other fields, and what they led to there: the strongest 12 of 31 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Game Theory is a planet in Nousfarer

Its city has a tower for each of these 12 frameworks. Land, walk in, and climb one concept at a time.

Land on Game Theory

Related fields

Fields its ideas are linked to

More in Microeconomics

See where Game Theory sits on the Map of Human Knowledge