Andreas_Giger comments on Paper: Iterated Prisoner’s Dilemma contains strategies that dominate any evolutionary opponent - Less Wrong

27 Post author: mapnoterritory 02 June 2012 08:50PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (19)

You are viewing a single comment's thread. Show more comments above.

Comment author: Andreas_Giger 03 June 2012 03:47:49PM *  1 point [-]

However, after working out the math, it appears that the optimal strategy against this one is actually a very nice one.

Of course, the same as in a game of chicken where your opponent precommits to defecting.

In infinite IPD:

  • There are lots of probabilistic strategies your opponent can precommit to that prevent you from averaging CC (in this case: 3).
  • If your opponent accepts any probabilistic precommitment from you without precommiting himself, you can maximise your score beyond CC.
  • If you model your opponent as a probabilistic strategy, you accept any probabilistic precommitment from your opponent.

Point 2 may not be obvious, but follows straight from the payoff matrix.

Comment author: Kindly 03 June 2012 05:11:09PM 0 points [-]

Well, yes; I'm assuming that I know the strategy my opponent is playing, which assumes a precommitment. I'm just trying to explain the reasoning in the paper, without going into determinants and Markov chains and so on.