For years, “Tit for Tat” (or close relatives such as “Tit for Two Tats”) has been the most successful strategy in the game-theoretic competition known as the Iterated Prisoners’ Dilemma. Now, apparently, a group at Southampton in the UK has devised a group-oriented strategy that can defeat TfT in a standard sort of simulated competition. This might seem like mind-numbingly boring news of interest only to math geeks unless you happen to know that study of this problem was a critical part of developing the Mutually Assured Destruction strategy of nuclear deterrence and also has significant implications for economics. Robert Axelrod even wrote a couple of books about the IPD, covering these and other implications.

The Southampton success actually seems a little bit questionable to me, though. It’s not that the programs maintained history about past interactions, because that’s what the IPD is all about as opposed to the single-shot PD. It’s not that the programs “colluded” because I actually think it’s rather clever. My concern is with the fact that the Southampton strategy’s success might not be generalizable. While nobody has yet succeeded in devising an anti-TfT strategy that would not only win itself but also cause TfT’s performance to degrade relative to the overall average, it’s easy to imagine an anti-Southampton strategy that would have such an effect. What makes TfT interesting is not that it succeeds against naive opponents who know nothing about it, but that it can still succeed (i.e. perform better than average) even in the presence of determined opponents like Southampton that know everything about it. In other words, much as Deeper Blue was (IMO) created to be a dedicated Kasparov-beater without actually managing to unseat him as the world’s best chess player, Southampton seems to have succeeded in becoming a TfT-beater without actually being an optimal or even robust IPD strategy. It’s an interesting result, but it remains an anomaly and doesn’t really affect the various lessons that have been drawn from studying TfT’s success.