Thursday, January 26, 2006

Interesting quote over at Michael Covel's blog

Michael Covel had a great quote a few days ago in a blog post:

  • "I did find one thing that worked. In fact almost all technical analysis can be reduced to this one thing, though most people don't realize it: the distributions of returns are not normal; they are skewed and have "fat tails." In other words, markets do produce profitable trends."


This is a very powerful realization. In Jack Schwager's "Market Wizards" book series, quite a few of the trader's who were interviewed spoke about using option pricing models with fat tailed price distributions. In essence, these early options traders were betting on trends before the crowd knew how to price them in to the options.

However, I disagree with another assertion that was made in the same quotation:

  • "I was very, very skeptical that technical analysis had value. So I used the computers to check it out and what I learned was that there was, in fact, no useful reality there. Statistically and mathematically all these tools - stochastics, RSI chart patterns, Elliot Wave, and so on-just don't work. If you code any of these rigorously into a computer and test them they produce no statistical basis for making money; they're just wishful thinking."


I think "straw-man" statements like this one contain a lot of misdirection for traders and investors. Of course no single indicator is going to have any consistent predictive value, it is only by combining technical indicators that you can get a significant statistical edge. So maybe that guy did try every indicator. So what? Did he try combining them? There are many fundamental reasons why various indicators have value, consider this post I recently made regarding simple moving averages.

The market is a deterministic system. Price movement is simply caused by market and limit orders. There is a natural structure that emerges from those market and limit orders, and an understanding of that structure provides an edge to the aware trader. There are two ways to identify an edge. You can do it with an exhaustive "trial and error" search, or you can identify conditions that cause the edge to emerge and validate by testing.

6 Comments:

At 12:27 AM, Blogger Michael Taylor said...

Funny...I found the "technical indicators are worthless" comments irritating too. I think another question is how did he test those indicators? Just use them in their conventional OB/OS sense? Did he try as a trend strength indicator (the reverse of what their typically used for)? Or like you said...did he try combinations with other indicators or divergences from price? Blanket statements like that make me feel the person performed bias testing.

MT

 
At 1:54 AM, Anonymous Technicator.NET said...

I like your deep thoughts when it comes to investing. Technical Analysis is as important as Fundamental Analysis and it does matter. There's a lot of supply&demand support to how a lot of the indicators work. Even a CFA that I know from Merryll Lynch (La Jolla, CA) believes in technical analysis.

Technical Analysis isn't new nor old, but traders/investors are just beginning to understand how it could work. Experienced T/A traders know that is works and have made money out of it.

 
At 9:38 AM, Blogger Dave Johnson said...

Your post kind of stirred some ideas and I figured I would do a little backtesting on my blog. Just some basic stuff to stir some debate. I posted the results
of an initial crossover system in my latest post.
Thanks for presenting fresh ideas to the blog universe, it helps me keep the mind chugging.

 
At 10:39 AM, Blogger The Trading Digest.com said...

Covel's comments about "technical indicators are worthless" show that he 's another one of the "either or" crowd. The crowd that for some reason feels the need to beleive 100% in either fundamental or technical analysis and dismiss the other as worthless.

I've found in my own trading that once I started preselecting watchlists for my system based on fundamentals, my profits improved greatly.

 
At 9:02 PM, Blogger Brett Steenbarger, Ph.D. said...

I appreciate your thinking on this issue. What I find is that technical indicators, for the most part, are predictive only within certain portions of their ranges (usually the extremes). This suggests that the market is neither wholly deterministic, nor wholly random. Simple linear statistical models are not adequate, because they look for relationships that exist across the full range of values for the variables. Non-linear modeling (which you can think of as separate regression analyses for different portions of the ranges of predictor variables) can capture these relationships (see www.salford-systems.com for sophisticated analytic software in this vein). FWIW, I find the best predictive relationships among the least mined indicators, which is why I've home-brewed many of the indicators on my site. - Brett

 
At 10:42 PM, Anonymous trader123 said...

"identify conditions that cause the edge to emerge" - can you elaborate on this? How do you define "condition"? Thanks.

 

Post a Comment

<< Home