Hannes Rusch simulated random-payoff two-person games (http://www2.units.it/etica/2013_2/RUSCH.pdf) and found that 3.4% of them fit a technical definition of “Prisoner's Dilemma”. The payoffs were uniformly distributed integers in the interval [0-N] and

Created by torekp on 2014-05-25; known on 2014-06-01; judged right by torekp on 2014-06-01.

• torekp estimated 4% on 2014-05-25
• torekp said “This was a dumb estimate, given that I gave 70% chance that 1.7% <= truth <= 5.1%, to the nearest 0.1%, and I considered the middle of the distribution to be more likely than the ends.” on 2014-06-01
• torekp   judged this prediction right on 2014-06-01.
• hrusch said “Correct answer: For large N the fraction converges to 1/24 which is roughly 0.0417. (This is for PDs with T>R>P>S for both players, i.e., no indifferences.)” on 2014-07-12