Calculating an expected value — LessWrong