Random thought: What is the optimal PD strategy under imperfect information? — LessWrong