Extreme Value Theory and Jeff Smith

There is a intriguing story in today’s San Diego paper about one Jeff Smith, a resident of Escondido, California. His house was about two miles from the further advance of the recent wildfires here — and it was gutted by fire.

How did it happen? Probably wind-borne embers that were carried to his house, and then caused his shake-shingle roof to catch fire. He was downwind of the main fire, and his type of roof is highly flammable, the sort of thing that people in wildfire areas are strongly encouraged to replace.

Nevertheless, it’s an interesting example of extreme value theory at work: He is an outlier, with no-one around him catching fire, even those with a similar roof type, and yet his house burned. Some people are baffled at what happened, but in a sense it’s predictable — outlier cases like, while unhappy for those caught out, are part of dynamic, risky systems.


  1. It’s also possible he burned it down himself, a punk did it or the cause was entirely unrelated to the wildfires (you know, houses do burn down absent wildfires.)

  2. Sure, it’s possible, but do we really need to resort to explanations via conspiracy and criminality? He was downwind of a wildfire; neighbors and Smith both told reporters that embers were falling in the neighborhood; and Smith had a highly fire-prone roof. It’s much more interesting — and likely — that he is simply an outlier case in a dynamic system.

  3. Sometimes a black swan is just a black swan :-)

  4. Even if, all apologies to my friend NNT, I cordially dislike the whole black swan thing.

  5. I have to dispute too it’s “much more interesting — and likely — that he is simply an outlier case in a dynamic system”. What are the odds that at least one house *somewhere* within a few miles radius of the advance line of the fire will burn from unrelated causes? It’s not necessary to postulate conspiracy or criminality. Accidents happen.
    Note this is solvable – it should be clear from an insurance or other forensic investigation whether the fire started from an ember on the roof, or from inside the house.

  6. It’s a black swan or an EVT outlier or any other way you want to slice it – we don’t know the true underlying probability distribution of the likelihood of catching fire (though we can get an ok radial approximation based on where the flames are, the wind, roof materials, etc.), and regardless if it were to be completely normally distributed, there are outliers on every tail.
    Or for the contrarian, it could NOT be an outlier, and many of the homes that DIDN’T burn are outliers based on the underlying probability distribution.

  7. As an aside, NNT’s Black Swan book was the most pompous reads I’ve had in a long time. There probably are some good ideas in there, but man, not worth the extraction effort…

    My wife and I played a drinking game one recent evening reading random passages to each other — the one with the less pompous one (unanimous voting required), drank. Yes, we are probably meant for each other :)

    Some of you folks are thinking about this a bit too hard: maybe you should try the game too — It gets really good when you loosen up!

  8. What’s highly ironic is that by many accounts, NNT is even more pompous in person.
    Fooled by Randomness is just as pompous as the Black Swan, but it’s got more meat to chew on.
    I do wonder what the underlying probability distribution is of one of his readers going out of their way to punch him in the face for being such an an arrogant bastard.