That's correct: my approach doesn't generalize to unique/rare events. The 'naive' or frequentist approach seems to work for weather predictions, and creates a simple intuition that's easier IMO to explain to laymen than more general approaches.