A stock operator's thoughts and ideas about market principles
Stock Trader and
Economy Student
J.P. Janssen's website.The purpose of the site is to discuss ideas and principles about markets. More |
|
Sections Empirical Tests Market Models Cycles Practical Trading Mathematical Psychology Biology Philosophy Personal Development The Importance of ... Poetry Global Warming Complete Archive |
|
The Special Sections Relative Sizes Quotes The Domino Effect |
|
Recommended Books The Edu. of a Speculator 
Practical Speculation 
The Black Swan 
Fooled by Randomness 
Market Wizards 
Reminis. of a Stock Operator 
|
|
Recommended Websites Daily Speculations Science Daily CIA World Factbook Project Gutenberg LibriVox |
Apr/25/2007
Visual Hypothesis Testing and the Gorilla Suit
When you see a chart you will most likely spot some patterns. This is true even if the data is random. When you see the data in a big table, you'll most likely notice nothing of significance. If you test the data statistically something of interest might show up - be it that you reject or accept your belief.
However, a common mistake is to test blindly, i.e. to run lots of test until you find something that fits random variations in the data, but not the reality. In other words you are no better than those who look at graphs and find heads and shoulders, trend lines and double bottoms.
But then again, why is the brain created to find patterns in randomness? This, I believe, is because statisticians - and other graph looking persons - are the first ones who actually do look at randomness after millions of years of evolution. Naturally we are not created to look at charts, nor to look at data (which are just measured observations of reality), but to function in the real world. The task of vision is to spot non-randomness, which truly exists everywhere.
Imagine being blindfolded and taken to an unknown place. When you open your eyes again you'll really fast see whether you're indoors or outdoors, if people are around etc. Somehow the brain must test the impulses it gets from the visual nerves against lots and lots of pre saved images. There's no way molecules randomly get together and shape something that looks like a human, so why should the brain use time and energy to test for such an impossibility?
If it's a gorilla you're seeing, it's most likely a gorilla and not a man in a gorilla suit.
Well, it might be the latter actually, but then again your eyes cannot help you. Then it's your intelligence that comes into play. The sounds, the smell (?) and the behavior of the creature should let you make a pretty good understanding of what's going on.
The same thing applies into looking at charts. Your eyes are just too good at hypothesis testing, so when you think you see something; beware of the gorilla suit.
Caravaggio (http://the3500.wordpress.com/) comments:
Hi JP. I've seen quite a men in gorilla suits in the markets of late, but I'm confident I've also seen at least one real gorilla. That begs the question of whether some patterns really may be non-random, but sufficiently infrequent that they can never be identified before the fact, which makes them as good as random from the trader's perspective.
I think the idea about spotting patterns/stories that aren't really there is an interesting them. It's also the theme of a new book by Taleb (Black Swans), which may be of interest.
JP remembers:
A long time ago I used to make random graphs in Excel and did technical analysis on them.
Caravaggio:
Hah, there is actually a web site that has a random chart and a chart of a security and it asks viewers to vote on which is the real chart. Last time I looked more than half had voted for the random chart!
| < Newer Article | Older Article > |
