You ignored the 'multiple trial part'. Single-trial probability for life is small, but there were multiple independent trials being in run in many places in parallel. And it needs to success only ~once for the whole 'life starts event' to kickoff.