the black swan

THE BLACK SWAN
A Black Swan is…
… Something you didn’t expect which has a
strong impact.
(So the more you think you know the more vulnerable
you are because your confidence in predicting the
likelihood of a given event will make you more vulnerable
in the case where you don’t)
Mediocristan
• In Mediocristan everything is constrained by boundary
conditions: time, the limits of biological variation, the
limits of hourly compensation, etc.
• Random variation of attributes exists in Mediocristan,
and can be usefully described by Gaussian probability
models (e.g. the bell curve)
• Because overall constraints are in effect for all
occurrences no single data point will have any great
effect on the mean or average of the whole.
• Examples: height of people, calories eaten per day,
wages earned by cab drivers.
Extremistan
• Extremistan is the land of scalability: variation within distributions is
unconstrained and unpredictable.
• Generators of events produce distributions with very large or very
small extreme values, relatively frequently. And those extreme
values often affect the sum of attribute values in a sample
distribution + the mean value of such distributions.
• The probability of occurrence of extreme values varies greatly from
Gaussian models.
• In fact, many attribute value distributions in Extremistan do not fit
any known models well.
• Examples: booksales, wealth, website hits.
• Since extreme occurrences can greatly affect statistical properties
of distributions from Extremistan, it is hard to make reliable
inferences from sample data.
Two examples
Sample 100.000 Mexicans:
• Height (Mediocristan): most extreme
occurrence will move the average only
0.001%
• Wealth (Extremistan): most extreme
occurrence will move the average 467%
•
Getting Carlos Slim in your sample is a 3-4 sigma
event (less than 0.1% chance), but it will blow your
model of ‘mexican wealth’ wide open.
The Barbell Approach
• Extreme conservatism + extreme risk taking.
Taleb
makes
money
Taleb
makes
money
Taleb
makes
money
Taleb
makes
money
… if you can’t barbell:
1. Have respect for time and nondemonstrative
knowledge
2. Avoid optimization: learn to love redundancy
3. Avoid prediction of small-probability payoffs
4. Beware the ‘atypicality’ of remote events
5. Beware moral hazard with bonus payments
6. Avoid some risk metrics
7. Positive or negative black swans?
8. Do not confuse absence of volatility with absence of
risk
9. Beware presentations of risk numbers