10 sigma in the news again

Galton Box (demonstrates normal distribution)
Image via Wikipedia

Risks are magnified after a historically volatile event like Thursday. The 11.3% intraday move was a 10 sigma event (if you treat the daily range of SPY as a normal Gaussian distribution).

Per the excellent blog at http://www.aleph.se/andart/archives/2009/09/ten_sigma_numerics_and_finance.html, a 10 sigma event can be expected to occur  in 1x 10^23 days or approximately never.

And yet it did.

So it should be a foregone conclusion that the market is not a Gaussian normal distribution, and statistical models built in that assumption are flawed.

Because the markets have very fat tails (unusual events are not so unusual after all), it is wise to not push the limits suggested by ANY theoretical model, since the probability that the model is wrong will be far greater than the low probability event that they discount.

These are the ideas behind the risk management rule in the Smart Steps model .

Reblog this post [with Zemanta]
Tortoise Capital Management © 1996 Frontier Theme