Why the events that matter most are the ones nobody foresaw

Nassim Nicholas Taleb was an options trader before he was an author, and The Black Swan (2007) arrived the year before the financial crisis – which did the sales no harm.

A black swan has three marks: it is extremely rare, it has enormous consequences, and it appears obvious in hindsight. The name comes from the fact that Europeans held for centuries that all swans were white, until someone reached Australia. A thousand confirmations became worthless in a moment.

Two kinds of randomness

Taleb’s most useful distinction. In Mediocristan, things follow the normal distribution: height, weight, calorie intake. Add the tallest human alive to a sample of a thousand and the average barely moves.

In Extremistan, that does not hold. Wealth, book sales, city size, returns, epidemics. Add Jeff Bezos to the same sample and the average is useless. Here a handful of events account for nearly the whole result.

Two kinds of randomness

Mediocristan

Normally distributed

• Height, weight, calorie intake
• No single value shifts the average appreciably
• The tallest human alive, in a sample of a thousand, changes almost nothing
• Standard deviation and risk models work

Extremistan

A few decide everything

• Wealth, book sales, returns, epidemics
• One value can dominate the whole sample
• Add Bezos to the same sample and the average is useless
• The same tools give false confidence

The mistake, Taleb argues, is that we use Mediocristan’s tools – standard deviation, the normal distribution, risk models – on things that live in Extremistan. That produces a false confidence worse than no model at all.

The turkey problem

A turkey is fed every day for a thousand days. With each day it has more statistical evidence that humans are benevolent. Its confidence peaks the day before Thanksgiving.

The point is the limit of induction: that something has gone well for a long time is not evidence it will continue. Often it is the opposite – long stability can mean risk has been quietly accumulating.

Why we think we could have seen it

• The narrative fallacy. We cannot tolerate randomness, so we manufacture causes after the fact. After a market crash there is always an explanation on the evening news, and it did not exist that morning.
• Silent evidence. We study those who succeeded, and never see the graveyard of those who did the same and vanished. That makes us systematically overestimate how well recipes work.
• The ludic fallacy. Dice and roulette have known probabilities. Reality does not, and using the casino as a model for the world confuses risk with uncertainty.

What to do in practice

Taleb is clear that the answer is not to become a better forecaster. It is to arrange things so you can survive the unforeseen – and ideally profit from it:

• Avoid situations where a single event can take you out.
• Buffers and generous margins, even when they look wasteful.
• Be careful with debt and leverage, which remove the slack.
• Expose yourself to upside with limited downside, not the reverse.

The criticism

The book is polemical, repetitive and generous with contempt for economists, forecasters and other authors. That makes it tiring, and makes it easy to dismiss a point that is actually good.

Substantively, several people have noted that the ideas are not as new as the presentation suggests – Frank Knight distinguished risk from uncertainty in 1921, and Keynes wrote about the same thing. The concept is also nearly unfalsifiable as a claim: if something unexpected happens, it was a black swan; if it does not, we are still waiting. And it has been applied to anything that surprises anyone, which infuriates Taleb himself.

What holds up is the distinction between the two kinds of randomness, and the advice to build to survive rather than to build to be right.

The walk-through includes the criticism the book has met.

Main topics

• What a black swan is, and what it is not
• Mediocristan and Extremistan – two kinds of randomness
• The turkey problem and the limits of induction
• The narrative fallacy – inventing causes after the fact
• Silent evidence, and why we only ever see the winners
• The ludic fallacy – why dice are a poor model for the world
• What to do in practice when you cannot forecast
• What the criticism amounts to

Questions this episode answers

• What is a black swan?
• What is the difference between Mediocristan and Extremistan?
• What is the turkey problem?
• Why are forecasts so unreliable?
• What is the narrative fallacy?
• How do you prepare for something you cannot foresee?
• What is the criticism of The Black Swan?

Key points

• A black swan is rare, has enormous consequences, and looks obvious in hindsight.
• Some domains follow the normal distribution. Others do not, and there the average is meaningless.
• The turkey has a thousand days of evidence that humans are kind. Then Thanksgiving arrives.
• We construct explanations afterwards and believe we could have seen it coming. We could not.
• The aim is not to forecast better, but to arrange things so you can survive being wrong.

Sources and references

• Nassim Nicholas Taleb – The Black Swan (2007, expanded 2010)