fat-tailed domains
Areas in which extreme, rare events occur far more often or matter far more than a normal distribution would predict, so small-probability events can dominate outcomes
As defined in Nassim Taleb on the Financial Crisis 03/23/2009
Where it comes up · 1
Most important claims · 3 of 3Search all claims for “fat-tailed domains” →
- Smaller-probability events in fat-tailed domains can be far more consequential than their frequency suggests, so one 100-year flood matters more than ten 10-year floods, and what matters is probability times impact, i.e., the expectation of the event.Nassim Taleb on the Financial Crisis 03/23/2009
- Estimating rare events is inherently unreliable because we have little data on them by definition, so a claim like a once-in-10,000-years event relies on assumptions or models rather than personal experience, and the smaller the probability, the more the estimate depends on a priori specification.Nassim Taleb on the Financial Crisis 03/23/2009
- We have absolutely no idea how to estimate very small probabilities, yet they can be very consequential in fat-tailed domains, so models that produce small probabilities from theory are dangerous.Nassim Taleb on the Financial Crisis 03/23/2009