Game Theory and Nash Equilibrium

A handful of mathematicians turned the messy human business of strategic choice into geometry — and discovered a fixed point that nobody could escape.

Most of economics, for most of its history, studied a single actor facing a passive world: a farmer choosing how much grain to plant, a consumer deciding between apples and bread. The mathematics of that world is the calculus of one mind against indifferent prices. But the moment you put two minds in a room — two firms setting prices, two countries arming, two animals fighting over a watering hole — the calculus breaks. Each player's best move depends on what the other will do, and what the other will do depends on what they think you will do, and so on, into infinity.

For centuries this regress had no formal answer. Then, in the middle of the twentieth century, three men — John von Neumann, Oskar Morgenstern, and John Nash — cut through it. They did not solve the problem by making the regress shorter. They solved it by finding the points at which the regress stops moving.

Von Neumann and the founding of the field

The story properly begins in 1928, when the 24-year-old John von Neumann published a paper titled Zur Theorie der Gesellschaftsspiele — “On the Theory of Parlour Games.” In it he proved the minimax theorem: in any two-player game of pure conflict, where one player's gain is exactly the other's loss, there is a rational strategy for each side that guarantees the best possible outcome against an adversary who plays equally well. The strategies could be mixed — a probability distribution over moves rather than a single move — and the theorem said that no matter how clever your opponent, you have a defensible plan.

Sixteen years later, with the economist Oskar Morgenstern, von Neumann turned the result into a 600-page book: Theory of Games and Economic Behavior (1944). The book is grand and difficult. Its claim is grander still: that the proper foundation of economics is not calculus borrowed from physics, but a new mathematics of interactive decision — game theory. A game, in their sense, is any situation involving two or more agents whose outcomes depend on each other's choices. Markets are games. Wars are games. Marriages are games. Treaties, auctions, evolution — all games.

The book worked beautifully for zero-sum conflicts. But most of life is not zero-sum. Both sides can win, both can lose, the pie can grow or shrink. Von Neumann's minimax did not say what rational players should do in such cases. That gap stood open for six years.

The Prisoner's Dilemma

The cleanest illustration of the gap — and of why game theory is uncomfortable — is a story invented in 1950 by two RAND Corporation mathematicians, Merrill Flood and Melvin Dresher, and dressed up in a parable by Albert Tucker. Two suspects, picked up for a joint crime, are interrogated in separate rooms. Each is offered a deal: confess and implicate the other, or stay silent. The catch is that the payoffs depend on what both do.

Look at the matrix from Player A's side. If B cooperates, A gets 3 by cooperating and 5 by defecting — better to defect. If B defects, A gets 0 by cooperating and 1 by defecting — again better to defect. Defection beats cooperation no matter what B does. B reasons identically. Both defect. Both end up with 1, when by cooperating they could each have had 3.

This is the scandal at the heart of game theory. Rational individual reasoning produces an outcome that is, by every player's own standard, worse. Adam Smith's invisible hand does not always sweep in this direction. Sometimes rationality and welfare diverge, and there is no clever trick to unify them — not within the rules of the game itself.

Nash's idea

In 1950, a 21-year-old graduate student at Princeton named John Forbes Nash Jr. submitted a 27-page doctoral thesis. In it he defined, for any finite game with any number of players, an idea he called an equilibrium point — what we now call a Nash equilibrium.

A combination of strategies is in equilibrium if no single player, holding everyone else's strategy fixed, can do strictly better by changing their own.

It is a definition of stability, not of optimality. The Prisoner's Dilemma's mutual-defect outcome is a Nash equilibrium: neither prisoner can improve their lot by unilaterally switching to cooperation. The fact that both could improve by switching together is irrelevant; the definition rules out only unilateral deviation. Stability and goodness, Nash showed, are different things.

The deep result of the thesis was an existence proof. Using a topological theorem of Kakutani's about fixed points of set-valued maps, Nash proved that every finite game with any number of players has at least one equilibrium in mixed strategies. This was not obvious. Before Nash, you could write down games whose stable points nobody knew how to find. After Nash, you knew at least one was always there to look for.

Geometrically, the picture is this. Each player has a best-response function: given a guess about what the others will do, it returns the strategy that maximizes their payoff. A Nash equilibrium is a point at which all the best-response functions agree — a strategy profile that is each player's best response to itself.

What the idea unlocked

Nash's equilibrium was at first thought a curiosity. Within a generation it was the lingua franca of economic theory. Auctions are designed against it: the rules of a spectrum auction are chosen so that bidding your true valuation is a Nash equilibrium, and the U.S. Federal Communications Commission has raised tens of billions of dollars from auctions designed on exactly that principle. Antitrust regulators model oligopoly pricing as a Nash equilibrium between competitors. The Cold War's nuclear standoff was framed by Thomas Schelling, drawing on Nash, as an equilibrium of mutual threat.

The idea also escaped economics entirely. In 1973, the biologist John Maynard Smith introduced the evolutionarily stable strategy — a Nash equilibrium for populations, in which a behavior is stable because no mutant strategy can invade. Animal fights, mating displays, even the ratio of males to females in a population are explained as equilibria, with natural selection playing the role of the rational player. In modern artificial intelligence, training a generative adversarial network is the explicit search for a Nash equilibrium between two neural networks — one generating, one discriminating. The AlphaGo program reasoned about Nash equilibria in tree search. Mechanism design, which won the Nobel in 2007, is the art of choosing rules so the equilibrium they produce is socially desirable.

What ties all of this together is the unsettling lesson of the Prisoner's Dilemma. Equilibria exist; they are stable; they are not always good. A rational world is not automatically a happy one. Almost every interesting question in economics, political philosophy and social design — how to write laws, structure markets, build institutions, align AI — reduces in some form to: can we change the rules of the game so that its equilibrium is one we actually want to live in?

Nash himself, after his thesis, drifted into deeper mathematics, and then into a long episode of schizophrenia from which he eventually recovered. He received the Nobel Prize in Economics in 1994, forty-four years after the 27-page argument that turned strategic interaction into geometry. The thesis is still in print. It is one of the shortest, strangest, and most consequential documents in the history of social science.