Nash Equilibrium

Zero-sum games are pure conflict: what I win, you lose. But most of life isn't like that. Two firms advertising, two countries arming, two commuters choosing routes, two flatmates deciding whether to do the washing-up — their interests are partly aligned and partly opposed. These are non-zero-sum games, and they need a subtler idea of "solution" than a single value. The answer, and one of the most influential ideas of the twentieth century, is the Nash equilibrium — the notion that won John Nash a Nobel Prize and reshaped economics.

Two payoffs per cell

When both players can win or lose independently, one number per cell is not enough — we need a bimatrix, each cell holding a pair (\text{Rose's payoff},\ \text{Colin's payoff}). The most famous of all is the Prisoner's Dilemma. Two suspects, held separately, may each cooperate (stay silent, loyal to the other) or defect (betray). Higher numbers are better outcomes:

	Colin: Cooperate	Colin: Defect
Rose: Cooperate	(3, 3)	(0, 5)
Rose: Defect	(5, 0)	(1, 1) ◀ Nash

If they could trust each other, both would cooperate and score (3,3) — clearly the happiest outcome. Yet the logic of self-interest drags them somewhere else entirely.

What "equilibrium" means

A Nash equilibrium is a combination of strategies — one for each player — that is stable against second-guessing:

No player can improve their own payoff by unilaterally changing strategy while everyone else keeps theirs fixed.
Equivalently, each player's strategy is a best response to the others': given what they do, you are already doing the best you can.

It is a self-enforcing agreement: at a Nash equilibrium nobody has a private incentive to break ranks, so it tends to be where a system settles. Crucially, "no one wants to deviate alone" says nothing about whether a coordinated change would help everyone — and that gap is where the drama lives.

Dominant strategies drive the dilemma

Look at the game from Rose's side. If Colin cooperates, Rose gets 3 by cooperating but 5 by defecting — defect wins. If Colin defects, Rose gets 0 by cooperating but 1 by defecting — defect wins again. Defecting is better no matter what Colin does: it is a dominant strategy. By symmetry Colin reasons identically, so both defect and land on (1,1).

Check that (1,1) is the Nash equilibrium: from it, if Rose alone switches to cooperate she drops from 1 to 0 — worse. Same for Colin. Nobody gains by deviating alone, so it is stable. And yet both players would be strictly better off at (3,3). Rational self-interest has led them, with iron logic, to an outcome that is worse for everyone. That is the dilemma — and it explains arms races, overfishing, price wars and why your flat's washing-up piles up.

Reading it off the bimatrix

Best responses make the equilibrium leap out of the grid. In each column circle the row Rose prefers; in each row circle the column Colin prefers. Wherever both circles land in the same cell, neither player wants to move — that cell is the Nash equilibrium. Step through it:

Rose's circles both fall on the Defect row; Colin's both fall on the Defect column; they coincide only at (1,1). That doubly-circled cell is the Nash equilibrium — even though the un-circled (3,3) would leave both players better off.

Existence: there is always an equilibrium

The Prisoner's Dilemma has one equilibrium in pure strategies. Other games have none in pure strategies — matching pennies, for one — but Nash proved the general rescue:

Every finite game with any number of players has at least one Nash equilibrium, once mixed strategies (randomising over your choices) are allowed.
A game may have several equilibria — pure, mixed, better and worse — and selecting which one a real system reaches is a hard and separate question.

This generalises von Neumann's minimax theorem beyond the zero-sum world: where minimax gives one game a single value, Nash gives every game at least one stable resting point.

Because a Nash equilibrium need not be good for the group, we can measure exactly how bad it is: the price of anarchy is the ratio between the social cost at the worst equilibrium and the cost of the best possible coordinated outcome. In road traffic it is startlingly concrete — Braess's paradox shows that adding a road can make everyone's commute longer, because selfish route-choosing settles into a worse equilibrium. Quantifying the price of anarchy is a whole research field, and it tells designers of networks, markets and traffic systems how much a little central coordination (tolls, protocols, incentives) could be worth.

The seductive error is to assume that because rational players land on a Nash equilibrium, that outcome must be good. The Prisoner's Dilemma is the permanent warning: its unique equilibrium (1,1) is worse for both players than the non-equilibrium (3,3). A Nash equilibrium is merely stable — no one gains by deviating alone — not efficient. "It's an equilibrium" answers "where will this settle?", never "is this the best we could do?" Confusing the two is how you end up defending outcomes that leave everyone worse off.