He's leaving out a critical factor which is the replication time (in relation to the time used for the failure chance). The probabilities will be lower if you take it into account unless you have very long replication times.