11
Feb
2008
I’ve been thinking a lot about options this year and focusing my Internet searching on that topic a bit more than usual. I recently stumbled across a paper co-authored by Nassim Taleb, the author of the widely popular books “Black Swan” and “Fooled by Randomness”. In this paper, he rails against the Black Scholes Merton Option pricing model and explains why it wasn’t “invented” by the Nobel Prize winning researchers Black, Scholes, and Merton but rather “repackaged.”
The article also delves into the hidden risk of options and their pricing and I urge my readers to download the paper, its a fast read. It made me realize why traders like Victor Niederhoffer blow up when selling options, they rely on rational statistics applied to an irrational market. Most of the time selling options works well, and is very lucrative for traders and instituions. Its when the irrational market rears its ugly head and you get wild standard deviation moves in your assets, that the Black Scholes Merton’s weaknesses are exposed; it can’t adjust for financial outliers.
The problem with those outliers is that you never know when they’ll happen and most often the majority of traders and investors are exposed. So the future option trader question for today is to ask oneself, “what are the hidden risks in options and can a model be created to help better reflect those risks?”
5 Responses
dc
February 11th, 2008 at 2:29 pm
1tom
I’m not an expert nn guy like most on this board but from what i do know if you take the process called JENKINS which is a linear form of nn and aply it as if you were cleaning up a data download before you load it. Then the outliars would be considered as data flaws and totally eliminated and never considered as the true trend of the instrument that you are trading.
The JENKINS method works best only on linear
prices where trending is prevalent throughout
the price. When it comes to nonlinear or random or groupings then trying to smooth it out might not even get the right results you would want. “THE PRICE IS EVERYTHING”
Saying that i ran into a format that a perticular nn uses it’s where > they look 10 bars ahead with 5% gain reculculation points and 1% stop loss reculculation points. In other words they would automaticaly reculculate on real time data feed even minute data.The only reason i mentioned that is becuase it seems like a good format to practice for option players aspecially when using 1 - 3hour
data on the underlying stock,but you must have live feed conection.
dc
February 11th, 2008 at 8:47 pm
2hi
Thought i would add with better spelling this time that someone might use an alert setup 3% gain 1% loss of the underlying on 1-3 hour time frame and 5% gain 1% loss on longer time frame.At which time they can redo their nn for that underlying to predict the 10 bars ahead possible(hopfully 80%+) outcome.
Tom
February 12th, 2008 at 5:47 am
3dc: Rapidminer offers a “preprocessing” set of operators under the preprocessing > data > outlier directory.
I haven’t used them because I inspect my data visually (see my 9 steps to neural net modeling success).
Although I might use operator called “distancebasedoutlierdetection” which uses a k-nearest neighbor algorithm to find outliers and trim them from the data set.
Evelyn
February 19th, 2008 at 12:51 pm
4What I get from Taleb is that options at either of the tail ends of the bell curve are dramatically mispriced.
And it’s not just options. Anytime there are strong network effects in the system, dramatic positive or negative changes can happen.
Tom
February 21st, 2008 at 6:03 am
5Evelyn, these BSE are everywhere but you are right, they tend to be at the tail ends. I agree with his theory that we think we live in a normal Gaussian distributed world, when in fact we don’t. He calls it mediocristan vs extremistan.
RSS feed for comments on this post
Leave a reply
previous post: Changes To The Member Section
next post: Biofuels Deemed a Greenhouse Threat
to top of page...