Joe Carlsmith comments on On expected utility, part 2: Why it can be OK to predictably lose