Equilibrium is mutual best response
A combination of strategies is a Nash equilibrium when each player's strategy is a best response to the others'. Check it cell by cell: ask of each player whether switching alone would pay.
A Nash equilibrium is a combination of strategies in which no player gains by changing alone. Some games have several such cells and some have none, and for the games with none a player may mix strategies with chosen probabilities. The mixed equilibrium of a two-by-two game is found exactly, by making each player's mix leave the other indifferent.
Core ideas
The definition is one sentence. Everything else in the lesson is what it does and does not promise.
A combination of strategies is a Nash equilibrium when each player's strategy is a best response to the others'. Check it cell by cell: ask of each player whether switching alone would pay.
A player who mixes picks each pure strategy with a stated probability, and a mix is a best response only if every strategy used in it is itself a best response. That is why a mixing player must be indifferent between the strategies mixed.
The definition says nothing about whether the outcome is good for the pair. A game may have an equilibrium that both would swap for another cell, and the previous lesson already showed one.
The material
Three games, the third of which has no pure equilibrium.
A combination of strategies, one for each player, is a Nash equilibrium when no player can get a larger payoff by changing their own strategy while the others keep theirs. An equilibrium in pure strategies is a cell of the table; an equilibrium in mixed strategies gives each player a probability for each pure strategy.
Pure equilibria are the doubly marked cells of the previous lesson, and a game can have none, one, or several. Chicken has two. Two drivers head for each other on a narrow road and each can swerve or go straight. If both swerve nothing is lost. If one swerves and the other does not, the one who swerves gets -1 and the other gets 1. If neither swerves they crash and each gets -10.
The cells Swerve, Straight and Straight, Swerve are equilibria. In the first, the driver going straight gets 1 and would get 0 by swerving, and the driver who swerved gets -1 and would get -10 by going straight. Swerve, Swerve is not an equilibrium, because either driver gains by going straight. It is the one cell in which nobody is hurt, and equilibrium does not choose it.
Matching pennies is the opposite case. No cell is an equilibrium, because in every cell the player who loses the match would rather have chosen the other side. Nash's theorem says that every game with finitely many strategies has an equilibrium if mixed strategies are allowed. The theorem is stated here and not proved. What the lab does is compute the mix for a two-by-two game.
Let the column player swerve with probability q. The row player's expected payoff from swerving is 0·q − 1·(1 − q), which is q − 1. From going straight it is 1·q − 10·(1 − q), which is 11q − 10. A row player who mixes must be indifferent between the two, so the column player's probability is the one that makes them equal:
The column player swerves nine times in ten. The table is symmetric, so the row player does too. At that pair of mixes each driver expects -1/10, which is below the 0 each would get if both simply swerved, and the two crash with probability 1/10 · 1/10, one time in a hundred. The mixed equilibrium is an equilibrium in the full sense, since at it neither driver gains by changing alone, and it is worse for both than the cell it does not use.
The surprise is whose numbers fix whose mix. Each player's probability is chosen so that the other is indifferent. A player does not randomise to be unpredictable at the cost of expected payoff; they randomise because, at that mix, the opponent has no better strategy to move to.
Interactive
best responses, equilibria, dominance and efficiency, computed from the payoffs
Method
Pure cells first, then the mix.
These are the pure equilibria. There may be none, one or several.
Write the row player's expected payoff from each row as a line in q.
That is the mix that leaves the row player indifferent. If the answer is not strictly between 0 and 1, no completely mixed equilibrium exists.
Let the row player play the first strategy with probability p and make the column player indifferent.
Pure and mixed together are the equilibria. Chicken has three, matching pennies has one.
Let q be the chance the column driver swerves.
The same calculation for the row driver gives 9/10 again, since the table is symmetric. The lab prints this as the mixed equilibrium, with the two pure cells beside it, and it is the only completely mixed equilibrium the game has.
Given a two-by-two table, list the pure equilibria, write one player's expected payoffs as lines in the other's probability, solve for the point where they cross, and say whether that mix is an equilibrium by checking that neither player gains from a pure deviation.
Something to change about this lesson? Write it down, tick it off when it is handled, and export the list when you want to act on it. Kept in this browser only: nothing is sent anywhere, and nobody is notified.
No recommendations yet for this lesson.