I went and ran this another 100 times, so I could see what it would look like without the variance. The mean scores are:

A: 32.03
B: 28.53
C: 32.48
D: 24.94
E: 28.75
F: 29.62
G: 28.42
H: 26.12
I :26.06
J: 26.10
K: 36.15
L: 27.21
M: 25.14
N: 34.37
O: 31.06
P: 26.55
Q: 34.95
R: 32.93
S: 37.08
T: 26.43
U: 24.24

If you're interested, here's the code for the test (takes a day to run) and the raw output for my run (an inconvenient format, but it shows the stats for the matchups).

Sorry if this is a stupid question, but this tournament looked to me like a thinly disguised version of:

"Construct a robot that can read code and interpret what it means."

which is a Really Hard Problem.

Is that not a fair description? Was there some other way to approach the problem?

The only way I can see to go about constructing a GOOD entrant to this is to write something that can take as its input the code of the opponent and interpret what it will actually DO, that can recognize the equivalence between (say):

return DEFECT

and

if 1: return DEFECT return COOPERATE

and can interpret things like:

if oppcode == mycode return COOPERATE return DEFECT

And I have no idea how to go about doing that. From the fact that the winning entrants were all random, it seems safe to say that no entrants had any idea how to go about doing that either.

Am I missing something here?

T was supposed to do a bit more than it did, but it had some portability bugs so I hastily lobotomized it. All it's supposed to do now is simulate the opponent twice against an obfuscated defectbot, defect if it cooperates both times, otherwise play mimicbot. I didn't have the time to add some of the obvious safeguards. I'm not sure if K is exploiting me or just got lucky, but at a glance, what it might be doing is checking whether the passed-in bot can generate a perfect quine of itself, and cooperating only then. That would be pretty ingenious, since typically a quine chain will go "original -- functional copy -- identical copy -- identical copy", etc.