What I Learned from The Black Swan

Two kinds of randomness

When faced with a model that contains randomness, we must determine whether the situation belongs in what Taleb calls "Extremistan" or "Mediocristan". In Mediocristan, the Bell Curve fits data. In Extremistan, we get hockey-stick curves.

Beware the use of the bell curve.

The error matters more than the estimate

When making an estimate (or prediction), the error is often so high that the prediction is useless. Further, we often don't need to predict so much as prepare. We don't need to know what exactly will happen to know that something will happen.

Limit your downside and increase exposure to upside

Since in "Extremistan" the upsides and downsides are unlimited, we need to limit our exposure to the downside. Otherwise, sooner or later, we will run out of luck. Similarly, we want to be well exposed to fortuitous upsides because, sooner or later, you will get lucky.

Taleb suggests 80-90% of our money should be invested in the safest bet we can make. The remaining 10-20% should be spread out among the most bets with unlimited upside we can.

Silent Evidence

Most successes in winner-take-all scenarios are based on luck. Nobody wins forever.

If you can model success with a purely random process, beware narrative explanations.


There's plenty more in the book, though I'd need to reread it to get at it. Also, I'd like to build these ideas into exercises and transformations to myself. We'll see how that goes.