Is there a strategy better than Tit-for-Tat when playing the Iterated Prisoner’s Dilemma (IPD)?
The IPD is used to model everything from evolution to economics to diplomacy. It's a mathematical game that has some very interesting, well-worked out features. Each player in the game can either try to cooperate with the other player and get a sure, low payoff, or a player can risk trying to be greedy and get more reward or possibly be totally screwed if the other player is also greedy. Up until now it was thought that there were mathematical reasons why nothing could really do better than Tit-for-Tat (cooperate if the other player does, be greedy if the last time the other player was greedy).
This paper seems to show that there are strategies which allow Player 1 to limit Player's 2 payoffs, using "extortion" as a tactic. The strategy is based on finding a set of playing probabilities that zero the determinant of the payoff matrix and allow Player 1 to strictly limit Player 2's rewards. If Player 2 wants to get the maximum possible reward they have to start cooperating with Player 1. The author's suggest that it's likely this strategy is evolutionarily stable.
Interestingly this bit of math seems to lead to strategies that are basically "keep cooperating until you decide to be greedy, then keep being greedy until you decide to cooperate again". There is very little influence of the other player's actions on the probability of cooperation.
I am NOT well-versed in game theory. Anyone out there better able to comment than I? Is this truly novel? Does it have serious implications for all those models of diplomacy and other important things?