[atrifix]

[ QUOTE ]
Your TFT-1, TFT-2, ... scenario will not be played out by the rational person. Again, the rational person knows that if he defects on the next to last round, the TFT guy will defect on the last round, thus the next-to-last-round defectgor ends up not maximizing his utility, SO he will NOT defect on the next-to-last round.

[/ QUOTE ]
But in this case, the TFT player plays irrationally, which contradicts our assumption that both players are rational. The TFT player does not seek to maximize his utility on the last round, as he could do better by playing TFT-1. Both players are rational, so TFT will not be played. Now, by invoking CKR, we can also see that TFT-1 will not be played, and so on.

[ QUOTE ]
They both might defect on the last round, unless there will be multiple games played with multiple people, and they know you defected on the last round -- there will be retribution to pay -- as no game is really "in a bubble", and previous games will affect subsequent games.

[/ QUOTE ]
This is more like a denial of perfect information. If the payoffs, strategies, length of the game, etc. are known beforehand, then both players can employ backward induction.

As long as all the assumptions hold, you won't play against a variety of players (since strategies like TFT are irrational), you'll only play against ALL D, so you can't do any better than ALL D.
[ QUOTE ]
Again, maximization strategy is to cooperate -- and use TFT to communicate your strategy.

[/ QUOTE ]
No, the Pareto-optimal strategy is to cooperate. That is, we have a paradox that is similar to the one-shot Prisoner's Dilemma--both players, by acting rationally, end up in a situation that is worse for everyone.