Game theory is a mathematical framework that explains how people, firms, governments, and other decision makers choose when each outcome depends on choices made by others, in the context of economics. A basic game theory model identifies the players, their possible strategies, the payoffs from each combination of choices, and the information available when they act. Economists use concepts such as dominant strategy, Nash equilibrium, payoff matrix, zero-sum game, and Prisoner's Dilemma to predict strategic behavior. The idea exists because an ordinary choice can be evaluated alone, but a strategic choice cannot: the best move changes when another player can react.
What a game actually is
A game is a model of strategic interaction with defined players, available actions, rules, information, and payoffs. It does not have to be entertaining or competitive. It only requires that at least one player's best choice depends on what someone else chooses.
Consider two food trucks deciding where to park at lunch. Each can choose the office district or the college. A truck's sales depend on its own location and on the other truck's location. Choosing a location is therefore not a private optimization problem. Each owner has to form a belief about the rival.
Every useful game specifies several pieces:
- Players are the decision makers. They can be people, businesses, political parties, countries, computer programs, or even biological populations in evolutionary game theory.
- Strategies are complete plans for what a player can do. In a one-move game, a strategy may be a single action. In a game with several stages, it must say what the player would do after every possible earlier event.
- Payoffs rank the outcomes for each player. They may represent profit, votes, time, safety, satisfaction, or a combination of several goals.
- Information states what each player knows before choosing. A bidder may know their own value for an object but not anyone else's value.
- Timing and rules determine who acts when, which promises are binding, and what actions are allowed.
You and a classmate must choose separate topics for a presentation. You each prefer an easy topic, but duplicate topics receive a penalty. Your best choice depends on what you expect your classmate to choose. That dependence makes the situation a game.
A model strips away details to expose the strategic link. Good modeling does not mean pretending that people care only about money. If fairness, reputation, or fear changes a choice, it belongs in the payoff. The model becomes misleading only when an important motive or available action is left out.
How a payoff matrix works
A payoff matrix lists the outcome of every strategy combination in a simultaneous game. Each cell contains one payoff for each player, in a stated order. A player compares payoffs while holding the other player's action fixed, then marks the best responses.
The Prisoner's Dilemma is the standard example because it shows how individually sensible choices can produce a worse result for everyone. The names come from a police interrogation story, but the same structure can describe price competition, pollution, shared work, or an arms race.
| Player A's choice | Player B cooperates | Player B defects |
|---|---|---|
| Player A cooperates | 3, 3 | 0, 5 |
| Player A defects | 5, 0 | 1, 1 |
The first number in each cell is A's payoff and the second is B's. These invented points express preference rankings, not dollars. For A, if B cooperates, defecting gives 5 instead of 3. If B defects, defecting gives 1 instead of 0. Defection is better for A in both cases. The same comparison holds for B, so both defect and receive 1 each, even though mutual cooperation would give 3 each.
Choose one column and compare the row player's payoffs within that column.
Repeat the comparison in every column, then do the equivalent comparison for the column player across each row.
A cell marked as a best response for both players is a Nash equilibrium.
The numbers need only preserve order when the model is about pure choices. A payoff of 5 need not be five times as enjoyable as a payoff of 1. Exact numerical differences become important when probabilities, risk, or mixed strategies enter the analysis.
How dominant strategies and Nash equilibrium work
A dominant strategy gives a player at least as good a payoff against every action others might take, while a Nash equilibrium is a set of strategies in which no player can gain by changing alone. Dominance concerns one strategy; equilibrium concerns everyone's choices together.
In the matrix above, defection strictly dominates cooperation for each player because it produces a higher payoff in every comparison. The resulting pair, defect and defect, is also a Nash equilibrium. If A alone switches to cooperation, A falls from 1 to 0. The same is true for B.
Holding everyone else's equilibrium strategy fixed, player i cannot improve by replacing the chosen strategy with another one.
The star marks an equilibrium strategy. The notation means the strategies chosen by all players other than player . This formal statement captures a simple test: freeze everyone else's move and ask whether one person wants to change.
A Nash equilibrium does not require a dominant strategy. Imagine two friends who want to meet and can choose either a library or a park. Both prefer meeting to missing each other, but one slightly prefers the library and the other slightly prefers the park. Meeting at the library and meeting at the park can both be equilibria. Neither location is best regardless of the other person's choice.
Equilibrium means no profitable solo move. It does not mean that players are happy, that the outcome is fair, or that society should preserve it.
Some games have one equilibrium, some have several, and some have no equilibrium in pure strategies. Nash's result guarantees an equilibrium for every finite game once mixed strategies, which assign probabilities to actions, are allowed.
Nash equilibrium versus the best collective outcome
A Nash equilibrium is stable against one-sided changes, while a collectively efficient outcome leaves no way to make one player better off without making another worse off. Stability and efficiency answer different questions, so an equilibrium can waste resources or harm every player.
Ask whether any one player benefits by changing while all others stay put. The test concerns incentives at a particular outcome.
Ask whether another feasible outcome helps at least one player without hurting anyone. The test compares outcomes, not unilateral moves.
Mutual defection in the Prisoner's Dilemma is stable but inefficient. Both players could receive 3 instead of 1 through mutual cooperation. Yet neither can reach that better cell alone. A unilateral cooperator receives 0, so the move toward the shared improvement must be coordinated or supported by changed incentives.
Economics often studies institutions that change a game's payoffs or rules. A pollution tax can make pollution more expensive. A contract can punish broken promises. Property rights can determine who may use a scarce asset. Antitrust law can restrict coordination that helps firms by raising prices but harms buyers. These interventions do not ask players to ignore incentives. They alter the incentives.
This distinction also connects strategic models to the method for comparing costs and benefits. A payoff table predicts choices under specified preferences, while a social evaluation asks whose costs count, whose benefits count, and how alternatives should be compared.
How sequential games change choices
Sequential games allow later players to observe earlier actions, so a strategy must include responses to every possible history. Analysts solve many such games by backward induction: begin with the final decision, identify the rational response, then work backward to the first move.
Suppose a new firm can enter a market or stay out. If it enters, the established firm can fight with a price cut or accommodate the entrant. Assume accommodation gives the established firm a profit of 4 and fighting gives it 1. Entry gives the newcomer 3 if accommodated and a loss of 1 if fought. Staying out gives the newcomer 0 and the established firm 6.
Start at the end. After entry, the established firm prefers 4 from accommodation to 1 from fighting. The newcomer anticipates accommodation and therefore chooses entry, receiving 3 rather than 0. A threat to fight does not deter entry because carrying it out would hurt the firm making the threat.
This is the idea of a credible threat. Words alone do not change the solution. A threat affects behavior only if the threatened action will actually be attractive when the moment arrives. A firm might make a response credible by signing a binding contract or installing capacity that changes the later payoff, but those steps have costs.
Timing also creates first-mover and second-mover advantages, depending on the rules. Moving first can let a firm commit to capacity. Moving second can reveal useful information. No universal rule says that acting first is better. The payoff structure decides.
How repeated games make cooperation possible
Repeated interaction links today's action to tomorrow's treatment. Cooperation can become an equilibrium when players value future payoffs enough and can observe behavior, because a short-term gain from cheating may trigger a larger stream of future losses, punishment, or lost trust.
Return to the Prisoner's Dilemma payoffs. Cooperation gives 3 each period. A one-time defection against a cooperator gives 5, but suppose it triggers mutual defection forever after, producing 1 per later period. Let represent how much a player values the next period relative to the present, with a value between 0 and 1.
With reward , temptation , and punishment , cooperation can be sustained when .
The left side is the present value of cooperating forever. The right side is the value of cheating once and then receiving the punishment payoff. The calculation does not prove that real people always use this severe strategy. It shows exactly how the future can change current incentives.
Cooperation becomes harder when players expect the relationship to end soon, cannot identify who defected, or frequently mistake accidents for cheating. It becomes easier when actions are visible, reputations travel, and responses can distinguish deliberate violations from noise. A shop protects repeat business differently from a seller who can disappear after one transaction.
Repetition can support harmful cooperation too. Competing firms may wish to keep prices high. Because coordination among rivals can reduce competition, laws and enforcement matter. Game theory describes the incentive; it does not supply a moral approval.
How game theory shows up in auctions and pricing
Auctions and price competition are games because each firm's or bidder's best action depends on rivals' actions and private information. The selling rule changes the strategy: a bid that makes sense in one auction format can lose money or surrender surplus in another.
In a sealed-bid, second-price auction, the highest bidder wins but pays the second-highest bid. Under the standard private-value assumptions, bidding one's true value is a weakly dominant strategy. Suppose you value an item at 80. If the highest rival bid is below 80, bidding 80 wins and you pay that rival bid. If the rival bid is above 80, losing avoids paying more than the item is worth to you. Misstating your value cannot create a better outcome in either case and can create a worse one.
A first-price sealed-bid auction works differently. The highest bidder wins and pays their own bid. Bidding your full value would leave no surplus if you win, so bidders usually shade their bids below their values. The sensible amount depends on beliefs about rival valuations and risk. The auction's rule changes the game, which changes behavior and revenue.
Firms also make strategic price and quantity choices. If two sellers offer close substitutes, a price cut may attract customers but provoke a matching cut. If each firm's decision affects total supply, the combined decisions help determine the market price. The connection between these strategic choices and how prices and quantities settle in a market is central to the economics of industries with a small number of large firms.
Financial trading can contain similar interactions. A trader considers not only an asset's future cash flows but also what other traders know, expect, and may sell. Models of signaling, information, and strategic order placement complement the broader study of how saving reaches borrowers and investments.
How game theory shows up in work, law, and daily decisions
Game theory appears whenever rules connect one person's reward to another person's response. Workplace bargaining, legal settlements, traffic, shared chores, international agreements, and sports all contain strategic choices, although each setting requires its own players, information, timing, and payoffs.
Bargaining divides gains and costs
Bargaining is a game in which parties can accept a proposed division or continue negotiating, strike, sue, or walk away. A person's outside option matters because it sets what they receive if agreement fails. Patience matters because delay can be costly.
An employer and worker negotiating pay are not simply dividing a fixed amount. Hours, flexibility, training, safety, and future promotion may also enter the payoff. A credible alternative job changes the worker's position. A credible ability to leave a vacancy open changes the employer's. Claims without usable alternatives may have little strategic force.
Law changes the game before conflict begins
Legal rules change expected payoffs by assigning rights, setting damages, requiring disclosure, and enforcing agreements. A contract makes some promises credible because breach has a specified consequence. Settlement bargaining then depends on likely court outcomes, legal costs, delay, and private information.
A rule can also create unintended strategies. If a benefit disappears at a sharp income threshold, people may change reported hours or timing to avoid crossing it. The relevant economic question is not only what the rule commands, but how informed players will respond.
Shared resources create social dilemmas
A shared kitchen, group project, fishery, or climate agreement can reward each participant for taking a little more while spreading the cost across everyone. If all reason that way, the resource deteriorates or the work remains undone.
Four housemates share a kitchen. Cleaning takes effort, and everyone enjoys a clean room even if someone else does the work. A rota changes the game by making duties visible and future responses predictable. It does not create generosity; it connects effort to accountability.
Real groups address these dilemmas with monitoring, repeated contact, social norms, deposits, access limits, and graduated penalties. Each device changes information or payoffs. The best design depends on the cost of enforcement and the chance of honest mistakes.
Sports reveal strategic unpredictability
A penalty taker choosing left or right and a goalkeeper choosing where to dive form a simultaneous game. If either side becomes predictable, the other can respond. Randomization can therefore be purposeful. Similar logic governs tennis serves, pitches in baseball, and play calling.
What zero-sum and non-zero-sum games actually are
A zero-sum game has a fixed total payoff, so one player's gain equals another's loss. A non-zero-sum game allows the total payoff to change, which creates room for mutual gain or mutual harm. Most market exchange and cooperation problems are non-zero-sum.
If two players wager 10 on a match and no fee is taken, the winner's gain is the loser's loss. The payoffs sum to zero after measuring each person's change in wealth. Chess is commonly modeled as zero-sum because a win for one side is a loss for the other, with a draw between them.
Trade is different. If one person values a book at 4 and another values it at 10, a voluntary sale at 7 can benefit both relative to no sale. The seller receives more than their value and the buyer pays less than theirs. Their interests still conflict over the price, but agreement can create total gains.
More payoff for one player necessarily means less for another. Strategy concerns how the fixed total is distributed.
Choices can enlarge or shrink total payoff. Strategy concerns both creating gains and dividing them.
Calling every negotiation zero-sum hides possible trades across issues. Two people may value money, timing, risk, and control differently. An agreement can exploit those differences while leaving a genuine conflict about distribution.
How mixed strategies work
A mixed strategy assigns a probability to each available action. Players randomize when predictable behavior can be exploited and no single action is always best. At equilibrium, the probabilities make an opponent indifferent among the pure actions that the opponent uses.
Matching Pennies gives a clean example. Each player secretly shows heads or tails. Player A wins if the coins match; Player B wins if they differ. Any fixed choice can be exploited. If A always chooses heads, B always chooses tails. If B always chooses tails, A switches to tails.
The equilibrium is for each player to choose heads with probability and tails with probability . This does not mean alternating mechanically, because an observable pattern is predictable. It means choosing so that the next action cannot be inferred from the previous sequence.
If an action produces payoff 4 with probability and payoff 0 with probability , its expected payoff is .
Mixed strategies describe frequencies or deliberate randomization, not confusion. They are useful in security patrols, inspections, sports, and military planning because a defender who always checks the same place at the same time invites exploitation.
What complete and incomplete information actually mean
A game has complete information when every player knows the relevant payoff structure and available strategies, even if actions occur privately. It has incomplete information when a player is uncertain about another player's type, such as costs, preferences, quality, or willingness to take risk.
Complete information is often confused with perfect information. Perfect information concerns observation of earlier moves. Chess has perfect information because the board and past moves are visible. A simultaneous pricing game can have complete information but imperfect information because each firm chooses without seeing the current choice of the other.
Private information creates signaling and screening. A job applicant may use a qualification to signal ability. An insurer may offer contracts with different deductibles so customers sort themselves according to risk. A seller may provide a warranty to signal product quality. A signal works only if it is sufficiently more attractive or less costly for the type it is meant to identify.
Labels depend on the model. Calling a qualification a signal does not prove that it measures skill or that it is socially useful. The claim must be tested against costs, alternatives, and observed behavior.
Bayesian games represent uncertainty by assigning beliefs to possible types. Players choose strategies that are best responses given those beliefs, then update beliefs when evidence arrives. This framework helps explain auctions, insurance, hiring, bargaining, and financial markets where private knowledge shapes offers.
Four mistakes people make with game theory
Game theory is often misused by treating a model as a full description of life, assuming equilibrium is desirable, overlooking timing, or forcing every conflict into a two-player zero-sum frame. Each mistake removes a part of the mechanism that may determine the result.
1. Treating payoffs as money only
People may value fairness, identity, duty, status, time, and the welfare of others. Those motives do not refute game theory. They mean the payoff was specified too narrowly. A prediction based only on cash can fail when social or legal consequences dominate.
2. Reading equilibrium as a recommendation
An equilibrium predicts stability under a defined set of incentives. It does not certify fairness, efficiency, or legality. Mutual pollution, discrimination supported by expectations, and aggressive retaliation can all be stable in a model. Evaluation requires additional evidence and ethical standards.
3. Ignoring commitment and timing
A simultaneous matrix cannot represent every sequential choice. Moving first, observing a signal, signing a contract, or removing a future option can change later best responses. Draw a game tree when history matters.
4. Assuming one model fits every strategic problem
A one-time Prisoner's Dilemma cannot automatically explain a repeated workplace relationship. A zero-sum model cannot capture gains from trade. A complete-information model cannot answer a question driven by hidden quality. The useful model is the smallest one that retains the decisive incentives and information.
Predictions also depend on how accurately actual people reason, learn, and understand the rules. Behavioral economics tests where observed choices differ from simple assumptions. Laboratory and field evidence can then guide a richer model rather than turning a tidy solution into an unquestioned fact.
Game theory connects individual incentives to economic outcomes
Game theory explains how separate choices combine into prices, contracts, cooperation, conflict, and institutions. It links individual incentives to economics as a whole by asking what each player can do, what each knows, what each values, and how others can respond.
The method is practical. When a decision feels strategic, write down the players and their possible actions. State the timing. Record what each player knows. Rank the outcomes, then search for best responses and credible later actions. If the result looks wasteful, ask which rule, contract, or repeated consequence could change the payoffs.
This approach belongs beside the broader explanations of economic choices and systems, because markets are built from people reacting to prices and to one another. Notice the next queue, negotiation, online auction, group assignment, or price change you encounter. Identify one response it invites. That response is where the game begins.
The takeaway: A strategic outcome cannot be understood by studying one choice in isolation. Map the other players' responses, and the incentives behind a puzzling result often become visible.