Nash Equilibrium

Nash equilibrium is the central solution concept in non-cooperative game theory and business and economics theory, defining a state of strategic rest in an n-player interaction in which each participant's chosen strategy is an optimal best response to the strategies actually chosen by the remaining n − 1 participants. Formulated by American mathematician John Forbes Nash Jr. in his 27-page Princeton doctoral dissertation and published in Proceedings of the National Academy of Sciences (1950) and Annals of Mathematics (1951), the concept generalized Antoine Augustin Cournot's (1838) duopoly model and John von Neumann's (1928) two-player zero-sum minimax theorem to arbitrary finite multi-player, non-zero-sum games. At a Nash equilibrium, expectations are mutually consistent and self-enforcing: if every player correctly anticipates the equilibrium profile, no single player has any incentive to deviate unilaterally.

Pure-Strategy and Mixed-Strategy Nash Equilibrium

Consider an n-player normal-form game G = ⟨N, (S_i)_i∈N, (u_i)_i∈N⟩, where N = {1, …, n} is the set of players, S_i is player i's set of pure strategies, and u_i: S₁ × … × S_n → ℝ is player i's payoff function. Let s = (s_i, s_−i) denote a strategy profile in which player i selects s_i ∈ S_i and all other players select the vector s_−i ∈ S_−i.

A strategy profile s^* = (s₁^*, …, s_n^*) is a pure-strategy Nash equilibrium if and only if for every player i ∈ N:

If the inequality holds strictly (u_i(s_i^*, s_−i^*) > u_i(s_i, s_−i^*) for all s_i ≠ s_i^*), then s^* is a strict Nash equilibrium.

In games where players randomize via probability distributions σ_i ∈ Δ(S_i), a mixed-strategy profile σ^* = (σ₁^*, …, σ_n^*) is a mixed-strategy Nash equilibrium if for every player i ∈ N and every alternative mixed strategy σ_i ∈ Δ(S_i):

Best-Response Formulation and the Mixed-Strategy Indifference Principle

Define player i's best-response correspondence BR_i: Δ(S_−i) ⇉ Δ(S_i) as the set of strategies that maximize player i's expected payoff against σ_−i:

Combining the individual correspondences into the joint mapping BR(σ) = (BR₁(σ₋₁), …, BR_n(σ_−n)), a strategy profile σ^* is a Nash equilibrium if and only if it is a fixed point of the joint best-response correspondence:

Because u_i(σ_i, σ_−i^*) is linear in player i's own mixing probabilities σ_i(s_i), an operational theorem—the Support Indifference Principle—follows: a mixed strategy σ_i^* is a best response to σ_−i^* if and only if every pure strategy s_i assigned positive probability (σ_i^*(s_i) > 0) yields the same maximum expected payoff, while any pure strategy outside the support (σ_i^*(s_i) = 0) yields a weakly lower expected payoff. Thus, in a mixed-strategy Nash equilibrium, each player chooses their randomization probabilities not to make themselves indifferent, but to make their opponents indifferent among the pure strategies in the opponents' active support.

Many simple games (such as Matching Pennies or Rock–Paper–Scissors) possess zero pure-strategy Nash equilibria. Nash's Existence Theorem guarantees that every finite game (having a finite number of players n and a finite set of pure strategies S_i for each player) possesses at least one Nash equilibrium in mixed strategies.

Nash published two distinct proofs of this theorem:

1. The 1950 PNAS Proof (via Kakutani's Fixed-Point Theorem): Because each mixed-strategy simplex Δ(S_i) is non-empty, compact, and convex, and expected utility u_i(σ_i, σ_−i) is continuous and linear (hence quasi-concave) in σ_i, Berge's Maximum Theorem implies that BR(σ) is upper hemicontinuous with non-empty, closed, convex values. Kakutani's (1941) Fixed-Point Theorem guarantees σ^* ∈ BR(σ^*)—paralleling the fixed-point technique later used by Arrow and Debreu in general equilibrium theory.
2. The 1951 Annals of Mathematics Proof (via Brouwer's Fixed-Point Theorem): Nash constructed an explicit continuous single-valued mapping f: Δ(S) → Δ(S) based on unilateral payoff gains g_i(s_i, σ) = max(0, u_i(s_i, σ_−i) − u_i(σ)):

Applying Brouwer's Fixed-Point Theorem ensures a fixed point f(σ^*) = σ^*, where all unilateral gain terms g_i(s_i, σ^*) must equal zero.

Cournot Duopoly and Tragedy of the Commons

In economic models with continuous strategy spaces S_i ⊂ ℝ₊ (such as quantities or prices), Glicksberg's (1952) Debreu–Glicksberg–Fan Theorem proves that a pure-strategy Nash equilibrium exists whenever S_i is compact and convex and u_i(s_i, s_−i) is continuous in s and quasi-concave in s_i. In an interior equilibrium, best responses are obtained by setting the first-order partial derivatives to zero:

In Cournot competition, two firms with marginal cost c facing inverse demand P(Q) = a − b(q₁ + q₂) maximize Π_i = (a − b(q_i + q_j) − c)q_i, yielding downward-sloping linear reaction functions q_i^*(q_j) = (a − c − bq_j) / 2b whose intersection defines the Cournot–Nash equilibrium q₁^* = q₂^* = (a − c) / 3b. Similarly, in the tragedy of the commons, independent harvesters equate private marginal benefit to private cost while ignoring the negative depletion externality imposed on other users, driving aggregate Nash extraction above the socially optimal level.

Equilibrium Refinements, Correlated Equilibrium, and PPAD Complexity

Because games can possess multiple Nash equilibria (by Wilson's 1971 Oddness Theorem, almost all non-degenerate finite games possess a finite odd number of Nash equilibria), game theorists developed structural refinements: Subgame Perfect Equilibrium (Selten, 1965) for dynamic games, Trembling-Hand Perfect Equilibrium (Selten, 1975) requiring robustness against small action perturbations, Bayesian Nash Equilibrium (Harsanyi, 1967) for auction theory and mechanism design under information asymmetry, and Correlated Equilibrium (Robert Aumann, 1974), where players condition actions on a public signal (such as a traffic light). In algorithmic game theory, Constantinos Daskalakis, Paul Goldberg, and Christos Papadimitriou (2009) proved that computing a mixed-strategy Nash equilibrium in a general finite game is PPAD-complete, establishing that no polynomial-time algorithm is guaranteed to find a Nash equilibrium in complex multi-player games unless PPAD = P.

Is a Nash equilibrium always the best possible outcome for the players?

No. A Nash equilibrium only guarantees that no single player can improve their payoff by changing their strategy alone (unilateral deviation). In games with strategic externalities—most famously the Prisoner's Dilemma, price wars, arms races, and the tragedy of the commons—the unique Nash equilibrium yields a worse outcome for every player than if all players could coordinate a joint deviation through an enforceable agreement.

Why would a rational player ever randomize in a mixed-strategy Nash equilibrium if all pure strategies in the mix yield the same expected payoff?

If a player is indifferent among their active pure strategies, they suffer zero expected utility loss from randomizing. However, if the player failed to randomize at the equilibrium probabilities σ_i^*—for example, if a tennis server always served to the forehand side—their opponent's best response would shift to a pure counter-strategy that lowers the server's payoff. The equilibrium mixing probabilities are required to prevent the opponent from exploiting predictable play.

What is the difference between a dominant-strategy equilibrium and a Nash equilibrium?

In a dominant-strategy equilibrium, each player's chosen strategy is optimal regardless of what any other player does (u_i(s_i^*, s_−i) ≥ u_i(s_i, s_−i) for all s_−i). In a Nash equilibrium, each player's strategy is only required to be optimal against the specific equilibrium strategies s_−i^* actually played by the opponents. Therefore, every dominant-strategy equilibrium is automatically a Nash equilibrium, but most Nash equilibria (such as those in coordination games or Cournot competition) are not dominant-strategy equilibria.

1. Nash, J. F. (1950). "Equilibrium Points in n-Person Games." Proceedings of the National Academy of Sciences, 36(1), 48–49. https://doi.org/10.1073/pnas.36.1.48
2. Nash, J. F. (1951). "Non-Cooperative Games." Annals of Mathematics, 54(2), 286–295. https://doi.org/10.2307/1969529
3. Kakutani, S. (1941). "A Generalization of Brouwer's Fixed Point Theorem." Duke Mathematical Journal, 8(3), 457–459. https://doi.org/10.1215/S0012-7094-41-00838-4
4. Aumann, R. J. (1974). "Subjectivity and Correlation in Randomized Strategies." Journal of Mathematical Economics, 1(1), 67–96. https://doi.org/10.1016/0304-4068(74)90037-890037-8)
5. Daskalakis, C., Goldberg, P. W., & Papadimitriou, C. H. (2009). "The Complexity of Computing a Nash Equilibrium." SIAM Journal on Computing, 39(1), 195–259. https://doi.org/10.1137/070699652