That's of course what had to be expected and what makes the game interesting, and also why the "evolutionary" tournament was supposed to give different results that the round-robin tournament. If there was a strategy which was best in any circumstances, everybody would send it.

Very much so. One thing that I've noticed in my own tests is that the presence of lots of sucker ("always cooperate") and defectbot strategies in the initial pool ultimately tends to favor vengeful strategies. The defectbots drive the suckers to extinction, but then can't survive without them.

Betrayers (who cooperate for a while and then start defecting if the opponent has always cooperated) consistently do very poorly, as do strategies that add random defection to tit-for-tat.

If there are enough sucker strategies ("always cooperate") then they can keep each other alive.

And under certain circumstances (that I haven't mapped out yet), all strategies I've tried do worse than random.

And I just ran one test on my own code (incidentally using the "Friend or Foe" payoff matrix rather than the Axelrod matrix) in which Random and Sucker dominated initially until the other non-nice strategies were driven to extinction, at which point a vengeful strategy took over. And then another on the same initial population where a forgiving tit-for-tat took over at the same point. This appears to be incredibly sensitive to initial conditions and the behavior of the Random Number God ...