This seem to work because in the end it’s a markov chain. So the probability of the next step of a long answer being correct is greater than the probability of

This seem to work because in the end it’s a markov chain. So the probability of the next step of a long answer being correct is greater than the probability of jumping to a correct conclusion.