1 Foundations
Game theory is the study of strategic interaction, meaning situations in which each participant’s outcome depends partly on the actions of others. It provides a formal language for analyzing cooperation, conflict, negotiation, and coordination. The field combines mathematics, economics, and social science to explain how incentives shape decisions.
1.1 Basic concepts
At its core, game theory describes who is involved, what choices they have, and what each choice produces. These elements form the structure of a game and allow strategic behavior to be modeled in a precise way.
1.1.1 Players
Players are the decision-makers in a game. They may be individuals, firms, political actors, animals, or even computer programs. A model specifies the set of players and assumes each one can choose among available actions.
1.1.2 Strategies
A strategy is a complete plan of action for a player. It may describe a single move in a simple game or a contingent plan for responding to future events in a more complex setting. Strategies are the main objects compared when analyzing outcomes.
1.1.3 Payoffs
Payoffs represent the results of the game for each player. They are often expressed as numerical utilities, profits, or rewards. Higher payoffs indicate outcomes that a player prefers over alternatives.
1.2 Assumptions of rationality
Many game-theoretic models assume that players act rationally, meaning they choose actions that best serve their preferences given what they believe about the situation. This assumption is central, though it is often adapted in behavioral or experimental work.
1.2.1 Preferences
Preferences describe how a player ranks possible outcomes. They may be based on money, status, safety, fairness, or other concerns. Game theory usually assumes preferences are consistent enough to be represented by payoffs.
1.2.2 Beliefs and expectations
Beliefs concern what players think others will do. Expectations affect strategic choice because the value of an action depends on anticipated responses. In many games, correct reasoning about beliefs is as important as the actions themselves.
1.3 Types of strategic interaction
Games can be classified by whether players can cooperate, whether they act independently, and whether interests are fully aligned or opposed. These distinctions help determine the appropriate analytical tools.
1.3.1 Cooperative games
Cooperative games study situations in which players can form binding agreements or coalitions. The focus is often on how groups share gains from cooperation. Allocation and stability are central concerns.
1.3.2 Non-cooperative games
Non-cooperative games analyze behavior when players choose actions independently and agreements are not enforced directly. This is the most common framework in modern game theory. It emphasizes individual incentives and equilibrium outcomes.
1.3.3 Zero-sum and non-zero-sum games
In zero-sum games, one player’s gain equals another player’s loss. By contrast, non-zero-sum games allow for outcomes in which all players can gain or all can lose. Most real-world strategic settings are non-zero-sum.
2 Game representations
Game theory uses several formal representations to describe strategic situations. Each highlights different features, such as simultaneity, order of play, uncertainty, or repeated interaction.
2.1 Normal-form games
Normal-form games represent choices in a compact tabular structure. They are especially useful when players move at the same time or when the exact timing of actions is not central to the analysis.
2.1.1 Payoff matrices
A payoff matrix lists the actions available to players and the resulting payoffs for each combination of choices. This format makes it easy to see best responses and possible equilibria. It is widely used for simple two-player games.
2.1.2 Pure and mixed strategies
A pure strategy selects one action with certainty. A mixed strategy assigns probabilities across available actions. Mixed strategies are important when no single deterministic choice is optimal.
2.2 Extensive-form games
Extensive-form games represent strategic interaction as a sequence of moves. They are suited to settings where timing, order, and information matter.
2.2.1 Game trees
A game tree maps the possible paths of play from start to finish. Branches represent choices, and terminal nodes show final outcomes. The tree structure helps identify future consequences of present actions.
2.2.2 Information sets
Information sets indicate what a player knows when making a decision. If a player cannot distinguish between certain nodes in the game tree, those nodes are grouped into one information set. This captures imperfect information.
2.2.3 Sequential moves
Sequential moves occur when players act one after another rather than simultaneously. Later players may observe earlier choices, which can create opportunities for reaction, commitment, or deterrence.
2.3 Other representations
Beyond standard normal-form and extensive-form models, game theory includes specialized frameworks for uncertainty, adaptation, and long-run interaction.
2.3.1 Bayesian games
Bayesian games model situations in which players have private information. Each player may know something about their own type that others do not. This framework is essential for analyzing incomplete information.
2.3.2 Evolutionary games
Evolutionary games study how strategies spread through populations over time. Rather than assuming deliberate optimization by individuals, they examine how successful behaviors become more common through selection or imitation.
2.3.3 Repeated games
Repeated games involve playing the same basic game multiple times. Past actions can influence future behavior, allowing punishment, reputation, and cooperation to emerge more easily than in one-shot interaction.
3 Solution concepts
Solution concepts are the rules game theory uses to predict or evaluate outcomes. They identify which strategies are stable, credible, or mutually consistent.
3.1 Dominant strategies
A dominant strategy is one that gives a player the best result regardless of what others do. When such strategies exist, they provide a strong and simple prediction of behavior.
3.1.1 Strict dominance
A strategy strictly dominates another if it yields a higher payoff in every possible case. Strictly dominated choices are generally considered irrational because they are always inferior.
3.1.2 Weak dominance
A strategy weakly dominates another if it is at least as good in every case and better in some. Weak dominance is a less demanding criterion, though its use can be more subtle in analysis.
3.2 Nash equilibrium
A Nash equilibrium is a set of strategies in which no player can improve their payoff by changing their own choice alone. It is one of the central ideas in game theory.
3.2.1 Pure-strategy Nash equilibrium
A pure-strategy Nash equilibrium occurs when each player chooses a single action and no one benefits from deviating individually. It is often easy to interpret, though not every game has one.
3.2.2 Mixed-strategy Nash equilibrium
A mixed-strategy Nash equilibrium involves randomization over actions. Each player chooses probabilities so that opponents are indifferent among their relevant options. This concept is useful in games with no pure equilibrium.
3.2.3 Existence results
A key mathematical result is that many finite games have at least one Nash equilibrium, possibly in mixed strategies. Existence theorems give the concept broad theoretical importance, even when equilibrium may be difficult to compute.
3.3 Refinements of equilibrium
Refinements are more demanding solution concepts designed to rule out implausible equilibria. They are especially important in dynamic and imperfect-information settings.
3.3.1 Subgame perfect equilibrium
Subgame perfect equilibrium requires strategies to be optimal not only in the full game but also in every subgame. It rules out threats or promises that would not be credible if the relevant point were reached.
3.3.2 Perfect Bayesian equilibrium
Perfect Bayesian equilibrium combines sequential rationality with consistent beliefs. It is used in games where players update expectations after observing actions. The concept is central in signaling and incomplete-information models.
3.3.3 Correlated equilibrium
A correlated equilibrium allows players to condition their actions on signals from a shared random device. Unlike Nash equilibrium, it can permit coordination through external recommendation while still preserving individual incentives.
3.4 Cooperative solution concepts
Cooperative game theory asks how groups can divide gains from joint action. Its solution concepts focus on fairness, stability, and the distribution of surplus.
3.4.1 Core
The core is the set of allocations that no coalition can improve upon by breaking away and acting on its own. If an outcome lies in the core, no subgroup has an incentive to object collectively.
3.4.2 Bargaining solutions
Bargaining solutions describe how players may divide a surplus through negotiation. They often aim to capture fairness, disagreement points, or strategic bargaining power. Different models emphasize different principles.
3.4.3 Shapley value
The Shapley value assigns each player a payoff based on their average marginal contribution across all possible coalition orders. It is widely used as a measure of fair distribution in cooperative settings.
4 Major classes of games
Games are often grouped by the timing of moves, the amount of information available, and whether interaction is one-time or repeated. These categories help determine the most appropriate analytical methods.
4.1 Simultaneous-move games
In simultaneous-move games, players choose actions without observing each other’s current decisions. The strategic problem centers on anticipating what others will do.
4.1.1 Coordination games
Coordination games reward players for choosing matching or compatible actions. They often have multiple equilibria because several coordinated outcomes can be stable. The main challenge is selecting the same focal point.
4.1.2 Anti-coordination games
Anti-coordination games reward players for choosing different actions. Such games often model competition for scarce opportunities or attempts to avoid crowding. Mixed strategies are sometimes used when players want to avoid predictability.
4.1.3 Prisoner’s dilemma
The prisoner’s dilemma is a classic game in which individual incentives lead to a worse collective outcome. It illustrates why rational players may fail to cooperate even when cooperation would benefit everyone.
4.2 Sequential games
Sequential games involve moves made in order, allowing later players to respond to earlier choices. This structure highlights commitment and the strategic value of timing.
4.2.1 Commitment
Commitment occurs when a player can bind themselves to an action or policy before others move. By limiting future options, a player may influence rivals’ behavior and improve their own outcome.
4.2.2 Credible threats
A threat is credible only if carrying it out would be rational when the moment arrives. Non-credible threats are usually ignored in equilibrium analysis because players would not actually want to follow through.
4.2.3 Backward induction
Backward induction solves a sequential game by reasoning from the end of the game backward to the beginning. At each stage, players choose actions that are optimal given the future path of play.
4.3 Games with incomplete information
Incomplete-information games arise when some relevant facts are unknown to players. This uncertainty may concern preferences, capabilities, or intentions.
4.3.1 Types
A type is a hidden characteristic that affects a player’s payoffs or available actions. Other players form beliefs about these types and adjust their strategies accordingly.
4.3.2 Signaling
Signaling occurs when informed players take actions to reveal information about themselves. A signal is useful only if it is costly or structured enough to separate different types.
4.3.3 Screening
Screening is the reverse process, in which an uninformed player designs choices to induce others to reveal private information. Menus of contracts or options are common screening devices.
4.4 Repeated and dynamic games
Repeated and dynamic games examine how strategic interaction unfolds over time. Past behavior can affect reputation, punishment, and expectations about future play.
4.4.1 Finite repetition
In finitely repeated games, players know the interaction will end after a fixed number of rounds. This endpoint can weaken cooperation because future punishments lose force near the end.
4.4.2 Infinite repetition
Infinite repetition refers to games with no fixed terminal date or with an indefinite horizon. Because future interaction continues, players may sustain cooperation through the threat of ongoing retaliation.
4.4.3 Folk theorem
The folk theorem shows that, under suitable conditions, many outcomes can be sustained in repeated games if players value the future sufficiently. It demonstrates the wide range of equilibria possible in long-run interaction.
5 Core models and examples
Several canonical games are used to illustrate central strategic problems. These examples appear throughout the literature because they capture recurring patterns of behavior.
5.1 Prisoner’s dilemma
The prisoner’s dilemma is one of the most famous examples in game theory. It shows how individually rational choices may produce a collectively inferior result.
5.1.1 Variants
Variants of the prisoner’s dilemma modify payoffs, the number of players, or the repetition structure. These changes can alter whether cooperation is fragile or sustainable. The basic tension between self-interest and mutual benefit remains central.
5.1.2 Applications
The model is used to study pricing, public resources, arms control, and many everyday social dilemmas. It is especially useful when participants face incentives to benefit from others’ cooperation without contributing themselves.
5.2 Chicken game
The chicken game models a confrontation in which each player prefers the other to уступe, but both risk the worst outcome if neither yields. It captures escalation and brinkmanship.
5.2.1 Conflict and risk
This game is often used to analyze situations involving danger, reputation, and willingness to stand firm. The central issue is how much risk players are prepared to accept to avoid backing down.
5.2.2 Mixed equilibria
When no player wishes to choose the same aggressive action with certainty, mixed strategies can arise. Randomization reduces predictability and may stabilize behavior when direct confrontation is costly.
5.3 Battle of the sexes
The battle of the sexes is a coordination game in which players prefer to coordinate but disagree about which outcome is best. It combines harmony with conflict over the focal point.
5.3.1 Coordination problems
The game illustrates the difficulty of aligning actions when each player has a different preferred equilibrium. Success depends on communication, convention, or preexisting expectations.
5.3.2 Multiple equilibria
Multiple equilibria are a defining feature of this model. More than one stable outcome exists, and the main analytical question is which one will be selected.
5.4 Public goods and social dilemmas
Public goods games study situations where individual contributions create benefits shared by many. They are a standard way to examine collective action and underprovision problems.
5.4.1 Free-riding
Free-riding occurs when a person enjoys the benefits of a shared resource without contributing to its cost. This behavior can undermine cooperation because each participant has an incentive to let others pay.
5.4.2 Collective action
Collective action refers to joint efforts to produce a shared outcome. Game theory explains why such action can be difficult even when the group as a whole would benefit from coordination.
6 Applications
Game theory is applied in many disciplines because strategic dependence appears in markets, institutions, biological systems, and networks. Its tools help organize complex interactions into analyzable models.
6.1 Economics
Economics is one of the main fields in which game theory is used. It helps explain pricing, bargaining, market structure, incentives, and information problems.
6.1.1 Market competition
Firms in a market often choose prices, quantities, or product characteristics strategically. Game theory models how each firm reacts to rivals and how these reactions shape outcomes such as profit and market share.
6.1.2 Auctions
Auctions involve bidders competing for goods while holding private information about value. Game theory is used to design auction rules and predict bidding behavior under different formats.
6.1.3 Contract theory
Contract theory studies agreements between parties with different information or incentives. Game-theoretic models help analyze moral hazard, adverse selection, and optimal incentive schemes.
6.2 Political science
Political science uses game theory to study voting, coalition-building, legislative bargaining, and strategic negotiation. It offers a structured way to analyze institutional behavior.
6.2.1 Voting
Voting models examine how individual preferences aggregate into collective choices. Game theory can explain strategic voting, agenda-setting, and the effects of electoral rules.
6.2.2 Bargaining
Bargaining models analyze negotiations over policy, resources, or office. They focus on disagreement points, bargaining power, and the credibility of offers.
6.2.3 Coalition formation
Coalition formation studies how groups form alliances to achieve shared goals. The size and stability of coalitions depend on how benefits and responsibilities are distributed.
6.3 Biology
In biology, game theory is used to study behavior that affects reproductive success, survival, and resource competition. Evolutionary models are especially influential.
6.3.1 Evolutionarily stable strategies
An evolutionarily stable strategy is one that cannot be invaded by a rare alternative strategy. This concept helps explain why certain behavioral patterns persist in populations.
6.3.2 Animal conflict
Animal conflict models analyze contests over territory, mates, or food. Game theory can describe when animals escalate, retreat, bluff, or settle disputes without prolonged fighting.
6.4 Computer science
Computer science uses game theory to analyze algorithms, distributed systems, and strategic behavior in digital environments. It is especially relevant where participants may have different objectives.
6.4.1 Algorithmic game theory
Algorithmic game theory studies games in computational settings. It examines how strategic agents interact with algorithms and how efficient outcomes can be computed or approximated.
6.4.2 Mechanism design
Mechanism design asks how to create rules that produce desired outcomes even when participants act strategically. It is often described as reverse game theory because the rules are chosen to shape incentives.
6.4.3 Network games
Network games model behavior on graphs, such as routing, communication, or the spread of influence. Outcomes depend on both local interactions and the structure of connections.
7 Mathematical and methodological extensions
Game theory has developed several extensions that refine its assumptions and broaden its applications. These include work on preferences, learning, experimentation, and critique.
7.1 Utility and preference theory
Utility theory provides the mathematical basis for comparing outcomes. It allows preferences to be represented numerically and used in strategic analysis.
7.1.1 Expected utility
Expected utility theory evaluates uncertain outcomes by weighting payoffs by their probabilities. It is a standard framework for decisions under risk and a foundation for many game-theoretic models.
7.1.2 Risk and uncertainty
Risk refers to situations where probabilities are known or estimable, while uncertainty involves less precise knowledge. Players’ attitudes toward risk can significantly affect strategic choices.
7.2 Learning in games
Learning models examine how players adapt their behavior over time rather than choosing optimally from the outset. These approaches are useful when agents have limited information or computational capacity.
7.2.1 Best-response dynamics
Best-response dynamics describe a process in which players repeatedly choose the best reply to others’ previous actions. This can lead to convergence, cycles, or other long-run patterns depending on the game.
7.2.2 Reinforcement learning
Reinforcement learning models adaptation through experience and rewards. Players gradually favor actions that have performed well in the past, even without fully solving the game.
7.3 Experimental game theory
Experimental game theory uses controlled studies to observe how people actually behave in strategic settings. It tests theoretical predictions against real decision-making.
7.3.1 Laboratory experiments
Laboratory experiments isolate strategic variables under carefully designed conditions. They allow researchers to compare theory with observed behavior in a repeatable environment.
7.3.2 Behavioral deviations
Behavioral deviations are systematic differences between predicted and observed actions. Common examples include fairness concerns, mistake-making, and limited attention.
7.4 Limitations and critiques
Game theory is powerful, but its models simplify reality. Scholars often examine where those simplifications are useful and where they may misrepresent behavior.
7.4.1 Bounded rationality
Bounded rationality recognizes that real decision-makers have limited time, information, and computational ability. This can lead to heuristic behavior rather than fully optimized strategies.
7.4.2 Multiple equilibria
Multiple equilibria make prediction difficult because several outcomes may be internally consistent. In such cases, additional assumptions or empirical evidence may be needed to identify the likely result.
7.4.3 Modeling assumptions
Game-theoretic conclusions depend on assumptions about preferences, information, and available actions. Small changes in the model can sometimes produce very different predictions.
8 History and development
Game theory emerged from efforts to formalize conflict and decision-making, then expanded into a broad framework for strategic analysis. Its development involved contributions from mathematics, economics, and social science.
8.1 Early precursors
Early work laid the foundation for later formal theory by exploring choice under conflict and uncertainty. These precursors helped establish the idea that interaction could be studied mathematically.
8.1.1 John von Neumann
John von Neumann made foundational contributions to mathematical economics and strategic analysis. His work on games of strategy helped establish the field’s formal basis.
8.1.2 Oskar Morgenstern
Oskar Morgenstern coauthored a seminal work on strategic behavior and economic decision-making. His collaboration helped connect mathematical methods with economic theory.
8.2 Formalization of non-cooperative game theory
The mid-20th century saw the development of equilibrium-based analysis for strategic interaction among self-interested players. This transformed game theory into a central discipline.
8.2.1 John Nash
John Nash introduced the equilibrium concept that bears his name, showing that stable outcomes can be defined for strategic games. His work became a cornerstone of non-cooperative game theory.
8.2.2 John Harsanyi
John Harsanyi advanced the analysis of games with incomplete information. His framework for types and beliefs made it possible to model uncertainty about private information rigorously.
8.3 Later advances
Subsequent researchers refined the theory, expanded its applications, and explored more realistic models of strategic behavior. These developments made game theory more flexible and empirically relevant.
8.3.1 Reinhard Selten
Reinhard Selten contributed important refinements to equilibrium analysis, especially for dynamic games. His work helped distinguish credible outcomes from formally possible but implausible ones.
8.3.2 Thomas Schelling
Thomas Schelling emphasized strategy in bargaining, conflict, and coordination. His ideas highlighted focal points, commitment, and the role of expectation in shaping outcomes.
8.3.3 Modern research directions
Modern game theory includes computational methods, behavioral models, evolutionary dynamics, and applications to networks and institutions. Research continues to refine solution concepts and improve descriptive realism.