In the sandbox, if you change the payoff of both parties cheating to -5, the angelic cooperator seems to win in most simulations. In a way, this makes sense: i

In the sandbox, if you change the payoff of both parties cheating to -5, the angelic cooperator seems to win in most simulations. In a way, this makes sense: if the penalty when both parties cheat is harsher than the upside of swindling your partner, "trust at all costs" wins out! But on a societal level, I suppose this means that if you want to optimize for do-gooders, you should punish failure harshly, no matter which party is responsible...