I have been somewhat a fan of Nassim Nicholas Taleb’s black swan idea, although not without reservations. A black swan is by definition unpredictable, and yet the current financial meltdown was hardly a surprise to anyone who cared to look. I am therefore pleased to mention a new entry into our statistical bestiary, that of the dragon king. In his paper Dragon-Kings, Black Swans and the Prediction of Crises, Didier Sornette identifies statistical outliers to data sets that otherwise follow power law distributions. Such outliers appear to be the result of positive feedback mechanisms, and two interesting hypotheses may be drawn: that they are both more common than expected, and also predictable.
“Million-to-one chances crop up nine times out of ten.”
– Granny Weatherwax, Equal Rites, by Terry Pratchett