We measure the error by considering the probability of getting close to the true value by a given amount, and then we try to bound it, given e and t.