Game Theory

Game theory is the mathematics of interdependent decisions: situations in which what is best for you depends on what someone else does, and what they do depends on what they expect you to do.

10 frameworks, from 1928 to today · Applied Mathematics · Mathematics

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

Overview

Game theory is the mathematics of interdependent decisions: situations in which what is best for you depends on what someone else does, and what they do depends on what they expect you to 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. The field builds formal models of such situations—players, the choices available to them, the information they hold, and the payoffs each receives—and then asks what rational players would do, what outcomes are stable, and how rules can be designed so that self-interested behavior produces good results1. Its results are used in economics, political science, biology, computer science, and engineering, and several of its ideas have reshaped how those fields think about conflict and cooperation.

Zero-Sum Games and the Minimax Theorem

The mathematical study of games began with parlor games, where one player's gain is exactly the other's loss. Such games are called zero-sum: the payoffs of the players always add to a constant. In 1928 John von Neumann proved the minimax theorem for two-person zero-sum games: in any such game (with finitely many pure strategies, allowing players to randomize), there is a value that each player can guarantee—one player can ensure at least that value, the other can hold them to at most it—by choosing a suitable mixed strategy, a probability distribution over their available moves. The theorem turned a vague notion of "playing well" into a precise number and a computable strategy.

The practical meaning is easiest to see in a simple game like matching pennies, where one player wins if the coins match and the other wins if they differ2. No fixed choice is safe, but randomizing fifty-fifty guarantees an even split against any opponent. Von Neumann's result said this kind of guarantee always exists in zero-sum games. It was the first deep theorem of the field and set the template: model the strategic situation, then prove what rational play must look like.

General Games and the Nash Equilibrium

Von Neumann, with the economist Oskar Morgenstern, published Theory of Games and Economic Behavior in 1944, arguing that game theory could be the foundation of economic analysis. But most economic situations are not zero-sum: trade, bargaining, and cooperation can leave both parties better off. The book's framework handled two-person zero-sum games well and general games less so.

The decisive step came from John Nash, in a short 1950 paper and a longer 1951 one. Nash defined an equilibrium—now called Nash equilibrium—as a profile of strategies, one per player, such that no player can do better by unilaterally changing their own strategy given the others'13. He proved that every finite game has at least one such equilibrium, possibly using mixed strategies13. The idea is a consistency condition on beliefs and actions: each player's choice is a best response to what the others are doing, so no one has a reason to deviate.

A concrete case is the Prisoner's Dilemma, formalized in the early 1950s by Merrill Flood and Melvin Dresher at the RAND Corporation and named by Albert Tucker14. Two suspects each do better by confessing whatever the other does, yet both would be better off staying silent1. The equilibrium—both confess—is stable but collectively poor1. The example made clear that equilibrium analysis describes what rational self-interested players will do, not necessarily what is good for them, and it became the field's most famous cautionary model.

Nash's result made game theory a general theory of strategic interaction rather than a theory of contests, and equilibrium analysis became the field's central tool. Its limitation was equally clear: many games have multiple equilibria, and the theory alone does not say which one players will reach.

Repeated Games and the Folk Theorem

One answer to the Prisoner's Dilemma is that people meet again. In a repeated game, the same stage game is played over many periods, and players can condition today's choice on what happened before15. In the 1950s and 1960s, work associated with RAND and with researchers including Robert Aumann and Lloyd Shapley established that repetition can sustain cooperation: in the Prisoner's Dilemma played repeatedly, a strategy of cooperating until the other player defects and then punishing them can make cooperation an equilibrium, provided players do not discount the future too heavily.

This crystallized into the folk theorem, so called because it circulated orally before being proved in generality. It says, roughly, that in repeated games with sufficiently patient players, a wide range of outcomes—essentially any outcome that gives each player at least what they could guarantee themselves—can be sustained as equilibria6. The theorem shifted attention from what equilibrium is to how many equilibria there are and which institutions select among them. It also gave economics a formal language for reputation, retaliation, and long-run cooperation, and it remains a workhorse in the study of cartels, treaties, and relational contracts.

Extensive-Form Games and Refinement

Games played in sequence—where players move one after another and observe earlier moves—are modeled as extensive-form games, drawn as trees17. Here a new problem appeared: equilibria can rest on threats that a player would never carry out. In a sequential game, a player might "threaten" a costly retaliation that is not in their interest if the moment arrives; an equilibrium in which the other player believes the threat can be stable in the formal sense yet implausible.

Beginning in the 1960s and accelerating in the 1970s and 1980s, Reinhard Selten and others developed equilibrium refinements: additional criteria that rule out equilibria relying on non-credible threats. Selten's subgame perfect equilibrium requires that the strategy profile be an equilibrium in every subgame, so threats must be credible at every point8. Later refinements—sequential equilibrium, proposed by David Kreps and Robert Wilson in 1982, and trembling-hand perfection—extended the idea to games with imperfect information. The refinements program showed that "equilibrium" is not one concept but a family, and that choosing among equilibria requires assumptions about what players know and believe.

Incomplete Information and Bayesian Games

Classical models assumed everyone knew the rules and payoffs. Real conflicts rarely work that way: a bidder does not know rivals' valuations, a negotiator does not know the other side's resolve. In 1967–68 John Harsanyi showed how to handle this by introducing types: each player has private information drawn from a known probability distribution, and a Bayesian game treats nature as moving first to assign types9. Equilibrium then requires each type of each player to choose a best response given beliefs about others' types. Harsanyi's transformation made asymmetric-information situations tractable, and it underpins auction theory, signaling models, and the analysis of bargaining under private information.

Cooperative Game Theory and Bargaining

A parallel tradition asked not how individuals choose but how coalitions form and how the gains from cooperation are divided. In Theory of Games and Economic Behavior, von Neumann and Morgenstern introduced the characteristic function, assigning a value to every coalition of players, and the concept of the stable set1. Lloyd Shapley's 1953 value gave a principled way to divide a coalition's total payoff according to each player's marginal contribution, and the core—the set of allocations no coalition can improve on—became the central solution concept. In the 1950s and 1960s, John Nash's bargaining solution and later work by Ariel Rubinstein (1982) on alternating-offer bargaining connected cooperative division problems to strategic, noncooperative models. Cooperative game theory remains the standard language for cost-sharing, coalition formation, and fair division.

Evolutionary Game Theory

In the early 1970s, John Maynard Smith and George Price asked how animal behavior could be shaped without rational calculation. They introduced the evolutionarily stable strategy: a strategy that, if adopted by a population, cannot be invaded by a rare mutant10. Maynard Smith's 1973–74 papers, and his 1982 book Evolution and the Theory of Games, applied this to animal contests—the Hawk-Dove game, for instance, models fights over resources where escalation is costly. The key move was to replace rationality with selection: strategies spread because they do better, not because players reason. This gave biology a rigorous account of why organisms behave strategically, and it fed back into economics as a way to justify equilibrium without assuming hyper-rational agents.

Mechanism Design

If game theory predicts how players behave under given rules, mechanism design asks the inverse question: given a desired outcome, what rules produce it11? The field took shape in the 1960s and 1970s. William Vickrey's 1961 analysis of auctions showed how bidding rules shape outcomes; Leonid Hurwicz formalized the design problem in the 1960s and 1970s; and in 1972 Roger Myerson characterized incentive-compatible mechanisms, leading to the revelation principle—roughly, that any outcome achievable by some mechanism can be achieved by one in which players truthfully report their private information. The 2007 Nobel Prize to Hurwicz, Maskin, and Myerson recognized this line11. Mechanism design is now the theory behind auction design, spectrum allocation, matching markets, and the design of institutions that work despite self-interest and private information.

Behavioral Game Theory

From the 1980s onward, experimentalists found that real players deviate systematically from equilibrium predictions: they cooperate more than the Prisoner's Dilemma predicts, reject unfair offers in ultimatum games, and reason a limited number of steps ahead. Behavioral game theory, associated with researchers including Colin Camerer and Ernst Fehr, built models of fairness, reciprocity, and bounded reasoning—such as level-k models, in which players reason "what would a level-0 player do, and how should I respond"—to explain the data. It does not replace classical theory but supplements it with empirically grounded assumptions about preferences and reasoning.

Algorithmic Game Theory

The most recent major turn came with computing. When games are large—internet auctions, network routing, online markets—players are algorithms and computation is costly. Algorithmic game theory, emerging around 2000, studies the computational complexity of finding equilibria, the price of anarchy (how much efficiency is lost when self-interested agents play rather than a central planner), and the design of algorithms for strategic environments. It connects game theory to computer science and to the design of online platforms, and it has made questions of computation and information central to the field.

The Present

Game theory today is a family of approaches rather than a single doctrine. Noncooperative equilibrium analysis, in its Nash and refined forms, remains the core; Bayesian games handle private information; repeated-game and folk-theorem results underpin the study of long-run relationships; cooperative theory supplies fairness and coalition concepts; mechanism design turns the theory into a design discipline; evolutionary and behavioral work supply alternatives to strict rationality; and algorithmic game theory addresses computation and scale. Open questions persist: which equilibrium concept applies when several are available, how real people actually reason, how to design mechanisms that are robust to misspecification, and how to model strategic interaction among artificial agents at scale. The field's history is a sequence of widening the model—from zero-sum contests to general games, from one-shot to repeated play, from complete to incomplete information, from rational to evolved and boundedly rational players, and from analysis to design.

References

  1. Game theory (introduction, Overview and terminology, Equilibria, Nash, As a formal discipline (early–mid 20th century), Early, theoretic Nobel-winning work (c. 1950s–1980s), Prisoner's dilemma, Repeated, Representation of games, Extensive form, Characteristic function form, Von Neumann–Morgenstern stable set), Wikipedia
  2. Matching pennies, Wikipedia
  3. Nash equilibrium, Wikipedia
  4. Prisoner's dilemma, Wikipedia
  5. Repeated game, Wikipedia
  6. Folk theorem (game theory), Wikipedia
  7. Extensive-form game, Wikipedia
  8. Subgame perfect equilibrium, Wikipedia
  9. Bayesian game, Wikipedia
  10. Evolutionarily stable strategy, Wikipedia
  11. Mechanism design, 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

    Zero-Sum Games and Minimax Theorem

    What strategy guarantees the best worst-case outcome?

    Fly to its tower in Nousfarer

  2. 1944-Present

    Cooperative game theory

    How should coalitions divide the gains from cooperation?

    Fly to its tower in Nousfarer

  3. 1950-Present

    Nash equilibrium

    What outcome is stable when each player best responds to the others?

    Fly to its tower in Nousfarer

  4. 1955-Present

    Repeated game

    When can cooperation be sustained over many repeated interactions?

    Fly to its tower in Nousfarer

  5. 1961-Present

    Mechanism Design

    How should rules be designed so self-interested agents behave well?

    Fly to its tower in Nousfarer

  6. 1965-Present

    Extensive-Form Games and Equilibrium Refinement

    Which threats and promises are actually credible in sequence?

    Fly to its tower in Nousfarer

  7. 1967-Present

    Bayesian game

    How do players act when they lack key information about others?

    Fly to its tower in Nousfarer

  8. 1973-Present

    Evolutionary Game Theory

    Which behaviors survive and spread in a population?

    Fly to its tower in Nousfarer

  9. 1985-Present

    Behavioral Game Theory

    How do real people actually play strategic games?

    About Behavioral Game Theory
    Behavioral game theory emerged to answer a question that classical game theory had largely set aside: how do real people, with their cognitive limits and social emotions, actually make strategic decisions? While traditional frameworks assumed perfectly rational, self-interested players who could solve complex games by reasoning alone, behavioral game theory took the experimental laboratory—and the psychological laboratory—as its point of departure. It asked whether actual human play conforms to equilibrium predictions, and when it does not, what psychological principles might explain the gap. The framework's core contribution to applied mathematics is a set of empirically grounded models that replace idealized rationality with bounded rationality, and purely self-regarding preferences with social preferences.
    Core curriculum A first course in behavioral game theory typically builds from the simplest experimental games toward richer models of strategic thinking and learning:
    1. Social preferences and inequity aversion: People care not only about their own payoffs but also about fairness. Models such as Fehr and Schmidt's inequity aversion capture the tendency to reject unequal outcomes even at a cost to oneself.
    2. Reciprocity: Players often respond to perceived intentions, rewarding kindness and punishing unkindness, which is distinct from pure outcome-based fairness.
    3. Bounded rationality and cognitive hierarchy (level-k reasoning): Players are limited in how many steps of iterated reasoning they can perform. Level-k models assume a distribution of reasoning depths, from naive (level 0) to increasingly sophisticated (level 1, 2, ...), matching experimental data from games like the "beauty contest."
    4. Quantal response equilibrium (QRE): A statistical equilibrium concept in which players make errors proportional to payoff differences, replacing perfect best-response with a probabilistic choice model.
    5. Learning in games: Players adjust their strategies over time based on experience, using reinforcement learning, belief learning, or imitation rather than one-shot equilibrium reasoning.
    6. Prospect theory and framing effects: From cognitive psychology, these concepts explain how people evaluate gains and losses relative to a reference point (loss aversion) and how the description of a strategic situation can shift choices.
    7. Ultimatum game, Dictator game, Trust game, and Public goods game: These are the canonical experimental workhorses. The ultimatum game reveals how proposers anticipate rejection of unfair offers; the dictator game measures pure altruism; the trust game quantifies reciprocity and betrayal aversion; and public goods games study cooperation and free-riding in groups.
    8. Experimental game theory: The methodological backbone that ties the curriculum together, encompassing laboratory protocols, field experiments, and statistical methods for testing behavioral models against observed play.
    Intellectual history Behavioral game theory did not appear in a vacuum. Its experimental and psychological roots reach back to the Allais paradox (1953) and the Ellsberg paradox (1961), which showed that people systematically violate the independence axiom of expected utility theory. In the 1970s, psychologists Daniel Kahneman and Amos Tversky documented a suite of cognitive biases—including prospect theory (1979)—that challenged the rational-choice foundation of classical game theory. Meanwhile, Vernon Smith's experimental markets in the 1960s and 1970s demonstrated that economic behavior could be studied under controlled conditions, laying the groundwork for experimental economics.
    But the distinct framework of behavioral game theory took shape in the 1990s and early 2000s, precisely the period of its local dominance (1990–2026). Seminal work by Matthew Rabin (1993) incorporated reciprocity into game-theoretic models, showing that fairness concerns could be formalized and tested. Ernst Fehr and Klaus Schmidt (1999) introduced a tractable model of inequity aversion that explained results from ultimatum and public goods games. Colin Camerer's Behavioral Game Theory: Experiments in Strategic Interaction (2003) codified the emerging field, synthesizing experimental evidence with models of learning, cognitive hierarchy, and quantal response. The level-k model was formalized by Rosemarie Nagel (1995) in the beauty-contest game and by Dale Stahl and Paul Wilson (1995); the quantal response equilibrium was developed by Richard McKelvey and Thomas Palfrey (1995). These works, all within the local window, made behavioral game theory a systematic research program rather than a collection of anomalies.
    Behavioral game theory coexists with, rather than replaces, its peers. Classical and non-cooperative game theory remain central for normative and benchmark analysis; mechanism design and algorithmic game theory incorporate behavioral constraints as extensions; and evolutionary game theory offers a complementary account of how boundedly rational agents might arrive at equilibrium through selection pressures. The framework's enduring legacy is to have made the empirical adequacy of game-theoretic models a central concern, bridging mathematics, psychology, and experimental economics.
    The 14 concepts its tower teaches
    • Bounded rationality
    • Cognitive hierarchy (level-k reasoning)
    • Dictator game
    • Experimental game theory
    • Framing effects
    • Inequity aversion
    • Learning in games
    • Prospect theory
    • Public goods game
    • Quantal response equilibrium
    • Reciprocity
    • Social preferences
    • Trust game
    • Ultimatum game

    Fly to its tower in Nousfarer

  10. 2000-Present

    Algorithmic Game Theory

    How can strategic outcomes be computed for very large systems?

    Fly to its tower in Nousfarer

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.

Zero-Sum Games and Minimax TheoremCooperative game theoryNash equilibriumRepeated gameMechanism DesignExtensive-Form Games and Equilibrium …Bayesian 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 58 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 10 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 Applied Mathematics

See where Game Theory sits on the Map of Human Knowledge