This stuck out to me: > Even under some very tolerant assumptions, the expected payoff from playing on, for either player, was greater than the expected payoff

This stuck out to me: > Even under some very tolerant assumptions, the expected payoff from playing on, for either player, was greater than the expected payoff from accepting the repetition. Payoff, sure. But it's well-known that the marginal utility of money is not linear, which means people tend to value money differently based on how much of it they have (poor people value a dollar more than rich people). This indirectly means some level of risk aversion is actually an optimal choice. Turning down a gamble, even if the expected payout is positive, can be rational. This is well studied in economic theory: https://en.wikipedia.org/wiki/Von_Neumann%E2%80%93Morgenstern_utility_theorem#Automatic_consideration_of_risk_aversion https://en.wikipedia.org/wiki/Von_Neumann%E2%80%93Morgenster...