Bostrom really failed there. He simply did not explain the argument very well. Everything Bostrom said is clear to someone who already understands the argument or has experience with statistics, but it was not at all a good explanation for someone who doesn't have the necessary intuitions to understand why they are more likely to be a simulated agent.
