Last year, AlexMennen ran a prisoner's dilemma tournament with bots that could see each other's source code, which was dubbed a "program equilibrium" tournament. This year, I will be running a similar tournament. Here's how it's going to work: Anyone can submit a bot that plays the iterated PD against other bots. Bots can not only remember previous rounds, as in the standard iterated PD, but also run perfect simulations of their opponent before making a move. Please see the github repo for the full list of rules and a brief tutorial.
There are a few key differences this year:
1) The tournament is in Haskell rather than Scheme.
2) The time limit for each round is shorter (5 seconds rather than 10) but the penalty for not outputting Cooperate or Defect within the time limit has been reduced.
3) Bots cannot directly see each other's source code, but they can run their opponent, specifying the initial conditions of the simulation, and then observe the output.
All submissions should be emailed to pdtournament@gmail.com or PM'd to me here on LessWrong by September 15th, 2014. LW users with 50+ karma who want to participate but do not know Haskell can PM me with an algorithm/psuedocode, and I will translate it into a bot for them. (If there is a flood of such requests, I would appreciate some volunteers to help me out.)
Found something slightly amusing:
When I enter trollBot into the simulation tournament, it actually ends up doing better than any of the default bots during the first round, pretty consistently. It also results in a win for defectBot.
I'm puzzled as to why trollBot does as well as it does. Is it just a function of the particular players it's up against, or is that actually a viable strategy?
You actually cooperate in this case.
Quick analysis: you're going to defeat CooperateBot 500 points), lose against DefectBot (0 points), and tie against TitForTatBot (250 points from alternating D/C and C/D). Against RandomBot, you are also RandomBot, both of you scoring 225 on average.
When you simulate MirrorBot, the infinite recursion makes ver time out, so you cooperate. So MirrorBot cooperates against you as well (300 points). SmarterMirrorBot and JusticeBot both time out as well. SmarterMirrorBot can... (read more)