No, this isn't a problem of probability, it's a problem of "not enough data." Based on this tiny sample, here's the lottery analogy: The lottery will flip 4 c

No, this isn't a problem of probability, it's a problem of "not enough data." Based on this tiny sample, here's the lottery analogy: The lottery will flip 4 coins that have a 37.5% chance of being heads. If you roll 4 heads lots of people die. So, just using this very inadequate sample from a single declassified document, you could estimate that a misfired bomb has a 2% chance of detonation (37.5%^4). Obviously none of us know the real dataset, but my point is simply that it's not comforting that so many safety measures failed. (Which, as it happens, is also the opinion of the author of the original report in the article.)