Game Theory
Quick Definition
Game theory is the mathematical study of strategic interactions, situations where the outcome for each participant depends not just on their own decisions but on the decisions of others. It provides a framework for analyzing competition, cooperation, negotiation, and conflict in economics, business, politics, evolutionary biology, and military strategy.
What It Means
Most economic decisions are not made in isolation. A company setting prices must consider how competitors will respond. A negotiator must anticipate the other party's reaction to each offer. A country deciding whether to build nuclear weapons must consider other countries' responses. Game theory provides formal tools to analyze these interdependencies and predict rational behavior.
The field was formalized by John von Neumann and Oskar Morgenstern in 1944 and extended by John Nash (whose life inspired the film "A Beautiful Mind"), who won the Nobel Prize in Economics in 1994 for his concept of Nash equilibrium.
Game theory has taken on new relevance in the 2020s. Algorithmic pricing, where software sets prices in real time based on competitor data, has raised concerns about tacit collusion. A 2024 study published in the Journal of Political Economy found that AI pricing algorithms can learn to collude without explicit communication, maintaining supracompetitive prices through repeated interaction. The FTC and DOJ have scrutinized this practice, and in 2025 the DOJ filed its first criminal case against a pricing algorithm provider in the real estate rental market. The case settled in early 2026 with a consent decree, but the broader question of when algorithmic pricing crosses into antitrust violation remains unresolved.
Core Concepts
Players, Strategies, and Payoffs
Every game has three elements:
- Players: The decision-makers (firms, countries, individuals)
- Strategies: The choices available to each player
- Payoffs: The outcomes (profits, utilities) resulting from each combination of strategies
Nash Equilibrium
A Nash equilibrium is a set of strategies where no player can improve their outcome by unilaterally changing their strategy, given what others are doing. It is the stable "resting point" of a game, not necessarily the best outcome for all players, just the point no one has incentive to deviate from.
Key insight: Nash equilibria can be highly inefficient. Everyone doing what is individually rational can produce outcomes worse for everyone than if they cooperated.
The Prisoner's Dilemma: The Most Famous Game
Two suspects are arrested. Each can either confess (defect) or stay silent (cooperate):
| Prisoner B: Silent | Prisoner B: Confess | |
|---|---|---|
| Prisoner A: Silent | A: 1 year, B: 1 year | A: 10 years, B: goes free |
| Prisoner A: Confess | A: goes free, B: 10 years | A: 5 years, B: 5 years |
Nash equilibrium: Both confess, getting 5 years each. But if both stayed silent, they would each get only 1 year. Individual rationality produces a collectively worse outcome.
Real-world prisoner's dilemmas:
| Scenario | "Confess" | "Stay Silent" | Outcome |
|---|---|---|---|
| OPEC oil pricing | Overproduce (cheat on quota) | Honor quota | Members cheat; prices fall |
| Arms race | Build more weapons | Disarm | Both build; both less safe |
| Advertising | Advertise heavily | Advertise less | Both advertise; same market share at higher cost |
| Carbon emissions | Emit freely | Reduce emissions | Countries emit; climate degrades |
| Algorithmic pricing | Raise price | Cut price | Algorithms learn to maintain high prices |
The prisoner's dilemma explains why cooperation breaks down even when it would benefit everyone, and why institutions, repeated games, and enforcement mechanisms matter.
Types of Games
| Game Type | Description | Example |
|---|---|---|
| Zero-sum | One player's gain = other's loss | Chess, poker |
| Non-zero-sum | Gains/losses don't cancel out | Prisoner's dilemma, trade negotiations |
| Cooperative | Players can form binding agreements | Mergers; treaty negotiations |
| Non-cooperative | No binding agreements | Most competitive markets |
| Simultaneous | Players decide at the same time | Rock-paper-scissors; sealed bids |
| Sequential | Players take turns; earlier moves observed | Chess; negotiation rounds |
| Repeated | Same game played multiple times | Ongoing business relationships |
Dominant Strategies
A dominant strategy is one that produces the best outcome regardless of what other players do:
Example: Pricing game between two competitors:
| Firm B: High Price | Firm B: Low Price | |
|---|---|---|
| Firm A: High Price | A: $10M, B: $10M | A: $2M, B: $15M |
| Firm A: Low Price | A: $15M, B: $2M | A: $5M, B: $5M |
For both firms, setting a low price is dominant (better in every scenario). Nash equilibrium: both choose low price, earning $5M, even though both could earn $10M with high prices (cooperation impossible without enforcement).
This explains why industries with few competitors (an oligopoly) often engage in price wars that hurt all participants. It also explains why companies invest so heavily in brand differentiation and product differentiation: the more similar the products, the closer the game resembles a prisoner's dilemma where price cutting is the dominant strategy.
Repeated Games: When Cooperation Emerges
In one-shot prisoner's dilemmas, defection dominates. But in repeated games (same game played multiple times), cooperation can emerge:
- Players who expect to interact again have incentive to maintain cooperation
- A tit-for-tat strategy (cooperate initially, then mirror opponent's last move) proves highly effective
- This explains why long-term business relationships, industry associations, and trade agreements facilitate cooperation
Robert Axelrod's tournaments showed tit-for-tat dominated in repeated prisoner's dilemma competitions by being nice, retaliatory, forgiving, and clear.
The repeated game framework also explains why algorithmic pricing algorithms can learn to collude. When pricing software interacts repeatedly with competitors' algorithms, it can learn that deviating from a high-price equilibrium triggers retaliation, making cooperation the stable strategy. This is not a theoretical concern: the 2025 DOJ case against RealPage demonstrated that algorithmic coordination in rental pricing can produce outcomes resembling explicit collusion without any human communication.
Game Theory in Business and Finance
| Application | How Game Theory Applies |
|---|---|
| Oligopoly pricing | Firms must anticipate competitors' price reactions |
| Auction design | Optimal bidding strategy depends on others' valuations |
| M&A negotiations | Bidder must anticipate seller's reservation price and competing bids |
| Patent racing | Firms racing to patent an innovation; first-mover dynamics |
| Salary negotiation | Employer and employee as strategic players |
| Options pricing | Market maker sets prices knowing informed traders exist |
| Central bank credibility | Fed's inflation-fighting effectiveness depends on market expectations |
| Tariff retaliation | Countries must weigh retaliation costs against deterrence benefits |
| Crypto mining | Miners compete for block rewards; each hash is a move in a stochastic game |
Key Points to Remember
- Game theory studies strategic interactions where outcomes depend on multiple players' decisions
- Nash equilibrium is the stable outcome where no player benefits from unilaterally changing strategy
- The prisoner's dilemma shows how individual rationality produces collectively worse outcomes, explaining arms races, price wars, and emissions failures
- Dominant strategies produce best outcomes regardless of others' choices; most games lack them
- Repeated games enable cooperation. Tit-for-tat strategies sustain mutual benefit when parties expect ongoing interaction
- Applied in oligopoly pricing, auctions, negotiations, regulatory design, and international relations
- Algorithmic pricing has introduced a new frontier where AI agents can learn tacit collusion without explicit communication, raising unresolved antitrust questions
Common Mistakes to Avoid
- Assuming players are perfectly rational: Real people and firms are bounded rational. They make mistakes, misjudge payoffs, and act on emotion. Game theory predicts what rational agents would do, not what every real player will do.
- Confusing Nash equilibrium with optimal outcome: A Nash equilibrium is stable, not necessarily good. The prisoner's dilemma equilibrium (both confess) is a Nash equilibrium, but it is worse for both players than cooperation.
- Ignoring repeated game dynamics: In one-shot interactions, defection often dominates. In repeated interactions, cooperation can emerge. Applying one-shot logic to ongoing relationships leads to wrong predictions.
- Overlooking information asymmetries: Many real-world games involve hidden information (one player knows something the other does not). Basic game theory models often assume complete information, which rarely holds in practice.
Frequently Asked Questions
Q: What is a Nash equilibrium in plain English? A: A Nash equilibrium is a stable outcome where every player is doing the best they can given what everyone else is doing. No one has a reason to change their strategy unilaterally, even if collectively the players might all be better off with different strategies. It is a "no regrets" equilibrium where each player is playing a best response to the others.
Q: What is the difference between game theory and decision theory? A: Decision theory analyzes optimal choices for a single decision-maker facing uncertainty (nature, probability). Game theory analyzes strategic choices when multiple rational agents interact. The "uncertainty" comes from other agents' choices, not random nature. When other players are involved, optimal strategy depends on anticipating their rational responses.
Q: How does game theory explain why OPEC cannot maintain oil prices? A: OPEC functions as a cartel where members agree to limit production to keep prices high. But each individual member has a prisoner's dilemma incentive to secretly overproduce: if others honor the quota, overproducing earns more revenue; if others cheat, you lose less by also cheating. Without binding enforcement, the dominant strategy is to cheat, and historically, OPEC members frequently violate quotas. The cartel only maintains discipline when Saudi Arabia (as the "swing producer") adjusts its own production to stabilize prices.
Q: Can AI pricing algorithms collude without communicating? A: Research suggests yes. Studies published in the Journal of Political Economy and by the OECD have shown that pricing algorithms using reinforcement learning can learn to maintain supracompetitive prices through repeated interaction, without any explicit agreement. The 2025 DOJ case against RealPage brought this theory into court, resulting in a 2026 consent decree. Regulators and courts are still working out how to apply existing antitrust law to algorithmic tacit collusion. See the FTC's algorithmic pricing guidance for more.
