A strategy used in a repeated game to secure an outcome which is not a Nash equilibrium for a single play of the game. Consider the prisoners’ dilemma. The outcome {Don’t confess, Don’t confess} is not a Nash equilibrium if the game is played only once. If the game is repeated an infinite number of times the strategy {Play Don’t confess if opponent played Don’t confess in previous round, Play Confess in every future play of the game if opponent ever plays Confess} is a punishment strategy that secures the equilibrium {Don’t confess, Don’t confess} if played by both players. The strategy rewards good behaviour (the play of {Don’t confess}) by responding in like manner and ensuring a high pay-off, but punishes bad behaviour (the play of {Confess}) by driving the pay-off down to the Nash equilibrium level.