I have quite a few friends at Morgan Stanley and many of their quantitative people are top-notch. So this has nothing to do with them. I do find the financial

I have quite a few friends at Morgan Stanley and many of their quantitative people are top-notch. So this has nothing to do with them. I do find the financial industry's underplaying of risk, in general, to be pretty hilarious. I remember when Goldman blamed a hedge-fund blow-up on three 25-sigma events happening in a week. That was in the summer of 2007. It elicited a good chuckle. If you sample from a Gaussian distribution, 25-sigma is about 1-in-10^137. That's 457 bits of improbability-- if the distribution is Gaussian. Three in a row? That's 1371 bits-- comparable to drawing 65 spade royal-flushes in a row in five-card poker. The mortality risk of a 25-year-old per Planck-time (5*10^-44 sec) is a mere 179 bits. Likely conclusion: financial events are not Gaussian, and existing models extrapolate poorly to new circumstances (sometimes this is out of an intentional desire to undercommunicate risks; more often, it comes from the business side's willingness to put more faith in simple models than the mathematicians that built them). Of course, this has been known for 30 years.