Black Swan 101

[Dr. Pataphoros, SpD]

Also requires a large impact to be a Black Swan.

Winning $5 million in imaginary money is not a Black Swan.

Losing $20 trillion in real money because a very very unlikely event just happened, on the other hand, is.

LOL

True enough in some sense. I think, Padre Pataphoros, that you may be confusing the value of the concept by misapplication. I don't think anyone has argued that Black Swans mean that all predictive data is useless, or that it has any effect on the nature of scientific prediction. Rather, its a philosophical point which focuses on the unknowable, which, throughout history seems to have been where the largest changes have wandered into reality from.

  When I asked for a clear definition of Black Swan THAT is the element that I was missing to make all of this make sense.

...


...  ...


Hm.  I no longer have a point to argue.  Carry on.

[Bebek Sincap Ratatosk]

Also requires a large impact to be a Black Swan.

Winning $5 million in imaginary money is not a Black Swan.

Losing $20 trillion in real money because a very very unlikely event just happened, on the other hand, is.

LOL

True enough in some sense. I think, Padre Pataphoros, that you may be confusing the value of the concept by misapplication. I don't think anyone has argued that Black Swans mean that all predictive data is useless, or that it has any effect on the nature of scientific prediction. Rather, its a philosophical point which focuses on the unknowable, which, throughout history seems to have been where the largest changes have wandered into reality from.

  When I asked for a clear definition of Black Swan THAT is the element that I was missing to make all of this make sense.

...


...  ...


Hm.  I no longer have a point to argue.  Carry on.

Then...

LET'S GET DANZING!

http://www.youtube.com/watch?v=dlf6n7IzZYk

[Bu🤠ns]

cain, or anyone else who has read the book:  Would you kindly elaborate more on the differences between Extremistan and Mediocristan including how scaleable knowledge vs. non-scaleable knowledge applies to both?  I get the idea i'm just having a hard time putting it in a simple "two dudes at a coffee break" explaination (if that's even possible).

[Cain]

cain, or anyone else who has read the book:  Would you kindly elaborate more on the differences between Extremistan and Mediocristan including how scaleable knowledge vs. non-scaleable knowledge applies to both?  I get the idea i'm just having a hard time putting it in a simple "two dudes at a coffee break" explaination (if that's even possible).

A scalable property is one that belongs in Extremistan - one that is subject to extremes.  A non-scalable property is one that belongs in Mediocristan.

For example, consider weight.  No matter how large a sample you have of people, when you study their weight on average, it will fall into a set range that the rest of the population resides in.  You don't get people who weigh 5000 tons, or 2 lb.  There are biological limits, and these can be tracked along the Bell curve.  Weight is a feature belonging to Mediocristan.

Now instead, let us consider income.  Income can be anything from nothing to the earnings of Bill Gates.  It is subject to high extremes and variations, because there are no physical limits on its growth.  It cannot be plotted by use of the Bell curve, and relying on the mean average is foolish, because the large variations can swing the "average" in one way or another (consider how GDP is used to calculate a countries wealth, especially in Saudi Arabia or China, where you have a small but fabulously wealthy upper class ruling over a destitute and numerous peasant class).  That is belonging to Extremistan.

[LMNO]

A simple test is this:  Will calculating the average mean of the group give a reasonable frame for all the people in that group?

Example:

1) weight.  210lb, 150lb, 190lb, 140lb, 280lb.  Average: 194lb.  All the given weights are close to this number.

2) income: $19k, $50k, $85k, $12k, $200M.  Average: $40,033,200  This number is completely unrelated to the individual incomes.

[Bu🤠ns]

so then do black swans happen in extremistan more often than they would in mediocristan due to the unpredictablitly of the limits? would it be a matter of basing knowledge on the end results rather than the processes?

[Dr. Pataphoros, SpD]

so then do black swans happen in extremistan more often than they would in mediocristan due to the unpredictablitly of the limits? would it be a matter of basing knowledge on the end results rather than the processes?

I think a Black Swan coming out of Mediocristan would be more surprising than from Extremistan.

[Cain]

According to the theory, it can't happen.  All "surprises" in Mediocristan fall within given and predictable range and thus are not very surprising at all.  Because the variables and possible results are known ahead of time, it is by definition impossible for a Black Swan to happen, because it needs to be unpredictable.

[Triple Zero]

what cain said, black swans only happen in extremistan. per definition, almost.

another point Taleb makes in his books is that there are a lot more phenomena that count as extremistan than we tend to think.

in fact, he argues that humans are biologically wired to discount extremistan phenomena and make predictions about them as if they were from mediocristan (not always, but a lot)

[Dr. Pataphoros, SpD]

According to the theory, it can't happen.  All "surprises" in Mediocristan fall within given and predictable range and thus are not very surprising at all.  Because the variables and possible results are known ahead of time, it is by definition impossible for a Black Swan to happen, because it needs to be unpredictable.

Didn't we just get done arguing that everything is essentially unpredictable?  Using the weight analogy, what happens when someone falls drastically outside the norm?  Like the (in)famous John Brower Minnoch, who weighed nearly 1,500 lbs. at one point.

[Triple Zero]

that example demonstrates it as well, assume this is the most extreme example you can find, it is less than 10x the amount of your average human.

compare to the annual income statistic, which can span multiple orders of magnitude.

mathematically, it's also a different distribution. weight is Gaussian, while annual income has an exponential-like (or worse) distribution, causing these effects.

[Dr. Pataphoros, SpD]

I see.  So, given the size of the sample there's a "sweet spot" where the variation within that sample is prime for unexpected results.  The overweight guy is way outside the mean, but he's still on the graph because the variation of possible wieghts is significantly smaller than the actual number of people in the human race.  Conversely, there are nearly as many salaries as there are people, so the spread is greater.

That sound about right?

[Cainad (dec.)]

You're getting there. Also consider the impact of each individual occurrence. Is the general rule (as determined by probability) or the exception more significant?

I thought a visual aid might be useful for presenting this topic, again using the "weight versus income" comparison. These examples require an understanding of what standard deviation is.

This is something from Mediocristan:


Notice that the guy who weighs 1,500 lbs, while very exceptional and quite rare (as indicated by his position on the standard deviation curve), ultimately is not very significant. On his own, he would not raise the average weight by more than a tiny bit, unless this graph were taken from a very small sample of people. Nor does he particularly skew the gene pool. The bulk of humanity still falls within a predictable range.

In this typical Mediocristan event, things that fall to the far end of the bell curve may be impressive or surprising, but they have little to no long-term influence. They are flukes, quirks, and generally ignorable.

Now let's look at something from Extremistan:


Not only does someone like Bill Gates practically fall off the chart because of his wealth, those few people who have annual incomes in the billions cause a graph of worldwide average income to become distorted. The number of people who actually fall into the "average" range is much less than standard probability dictates.

So we see that in Extremistan, the people and things which are at the tail ends of the probability curve are actually far more significant than the people and things which fall into the more predictable range. The extreme concentration of wealth in the hands of a small elite, and the subsequent enormous population of poor people, has far more impact than the so-called average annual income of the world's population.

(Note: even though it is from Extremistan, this is not an example of a true Black Swan because everybody already knows about it)

(Graphs not to scale, you pedantic spags)

[PeregrineBF]

Black swans can occur even in Mediocristan. Meteors, volcanoes, Outside Context Problems, etc. No matter how well you define your context there will always be outliers, and in at least some of the cases these outliers will be very, very big.
Either that, or someone comes around with a bottle of spray paint and starts decorating the swans.

[Verbal Mike]

No PBF, I think the point is that when a Black Swan shows up in Mediocristan, that means it was actually Extremistan. If you believe a certain volcano to be perfectly safe and dormant, its eruption is one hell of an Extremistani Black Swan.