Nassim Taleb
The Back Swan
Technical Terms
|
Nassim Taleb |
The Back Swan |
Technical Terms |
|
|
|
|
Some key terms used throughout The Black Swan... |
possumgolightly |
Data that is far removed from the mean. Some outliers are data problems - the result of a mistaken measurements, some indicate a different sort of reality - one that allows wild variations (black swans).
A curve used to represent data that clusters around the middle and that falls off equally on both sides. Reality that can be represented by a bell curve is either physically constrained (like the size of human heads) or is isolated from other occurrences (like coin tosses). In a Gaussian world the more measurements you make, the more closely your curve fits the ideal.
In Taleb's terms, Mediocristan is a Gaussian world. Things tend toward the middle. Outliers are unlikely. In contrast, the events of Extremistan do not fit neatly under a bell curve. Outliers are likely.
A measure of deviation from the mean of a bell curve. Sigma is also called standard deviation.
The higher the number, the greater the deviance; the less likely an event is to occur. For example, in the above diagram, a 3 sigma event has a 0.3% chance of happening.
When Extremistan is mistaken for Mediocristan, impossible sigmas can result. For example, if the stock market crash of 1987 were represented on a standard bell curve, it would be a 20 sigma event - which shouldn't happen during the lifetime of the universe.
A relationship between data where things get big (or small) in a hurry. In a power law reality, the value of the first place (of something being ranked) is x, the value of the second place is half the first (x/2) the value of the third place is one third the first (x/3), the value of the tenth place is x/10, etc.
The relationship is represented by a power law or Pareto distribution curve.
The types of reality controlled by power law relationships include...
Distribution of wealth (in 18th century Vilfredo Pareto noted that 20% of people have 80% of wealth - thus leading to the "80/20" rule)
Earthquake magnitude (a few big earthquakes do more damage than a lot of little earthquakes)
Size of cities (a few big cities have more people in total than a lot of little cities)
Book sales (a few blockbuster books have more sales in total than a lot of also-rans)
Word usage (linguist George Ziph observed that some words are used a lot and a lot of words aren't used much)
Stock market activity (one bad/good day can wipe out a lot of so-so days)
Web site/blog links (a few sites/blogs have most of the links)
In power law reality things are connected (even minimally - it doesn't matter). Feedback loops occur. This affects that which affects this which affects that and so on (in recursive Mandelbrotian fashion). Things start out equal then always, inevitably become unequal. Power law reality is inherently unfair.
Also, power law reality is "scale invariant" - meaning that the relationship holds in any subset of the data. This makes it fractal or Mandelbrotian process.
Extremistan is a power law place. The few that dominate the many are black swans. Problems (inevitably) happen when people confuse Gaussian reality for power law reality.
Refers to the long end of a power law or Pareto distribution curve (where a few occurrences of what ever is being measured are responsible for more than half the data - shown at the head of the curve). In web business, the term was first applied as a proper noun by Chris Anderson to describe the operation of companies like Amazon and NetFlix. Rather than sell a large quantity of a small number of items (from the head of the curve) they sell small quantities of a large number of items (from the tail). They mine the long tail.
In Extremistan, most of us labor unnoticed and unrecognized in the long tail - trying to figure out how to move to the head.