### Tuesday, December 27, 2005

## Technical analysis on trial

Despite a nice move in APN, my portfolio floundered today with the rest of the market, falling by about $2,800.

I recently came across an interesting survey article on the evidence for the effectiveness of technical analysis. Of the 92 modern studies of TA cited in the survey, 58 found positive results regarding technical analysis. Mind you, many of the papers suffered from methodological problems, including the failure to model transaction costs (commissions and bid/ask spreads) and possible data mining. Data mining (or data snooping as it’s sometimes called) is particularly important because if you take a large enough data set, you will always find some data patterns that appear non-random. To illustrate this point, think about what happens when you take the square root of 2. The resulting number (1.4142….) is an irrational number (i.e. the sequence of numbers after the decimal point never repeat). If you were to assign A=1, B=2,… Z=26, you would eventually find a complete sequence of numbers in the number 1.4142... that unerringly and in the proper order spell out the complete text of the bible (in mathematics, there is a world of difference between a very big number and an infinite one). Does that mean that there is something magical about the square root of 2? Of course not!

In the context of technical analysis, if you backtest enough patterns, you will find that some yield consistently profitable trades (and indeed, it is these patterns that are likely to become popular and be cited by adherents of technical analysis). Thus, it is crucial to test any ostensibly profitable trading rule using out-of-sample data (data different from the one used to identify the rule). When one does this, it often appears that trading rules that are profitable in one time period cease to be profitable in the following time period. Indeed, the authors of the article concluded that technical trading rules ceased to be profitable for U.S. stocks after the late 1980’s. Perhaps this is because their increased use neutralized their efficacy? It’s hard to say. However, Andrew Lo’s description of the stock market as an evolutionary mechanism in which the profitability of various trading rules is always changing appears to hold much truth.

Nevertheless, I am impressed that at least some of the studies covered by this survey article did isolate trading rules that yielded extremely large annualized returns (over 200 percent annualized in some cases). This may suggest that there are some profitable technical trading rules out there that still waiting to be discovered. Of course, the only problem is that once they are discovered and disseminated to the broader public, they may cease to exist. In the meantime, it seems like the only consistent way to make money using technical analysis is to write a newsletter on the subject!

p.s. Thank you to the person who left a comment telling me that the original version of this post actually used a rational number (2/7) and not an irrational one (egg on my face!!!)

I recently came across an interesting survey article on the evidence for the effectiveness of technical analysis. Of the 92 modern studies of TA cited in the survey, 58 found positive results regarding technical analysis. Mind you, many of the papers suffered from methodological problems, including the failure to model transaction costs (commissions and bid/ask spreads) and possible data mining. Data mining (or data snooping as it’s sometimes called) is particularly important because if you take a large enough data set, you will always find some data patterns that appear non-random. To illustrate this point, think about what happens when you take the square root of 2. The resulting number (1.4142….) is an irrational number (i.e. the sequence of numbers after the decimal point never repeat). If you were to assign A=1, B=2,… Z=26, you would eventually find a complete sequence of numbers in the number 1.4142... that unerringly and in the proper order spell out the complete text of the bible (in mathematics, there is a world of difference between a very big number and an infinite one). Does that mean that there is something magical about the square root of 2? Of course not!

In the context of technical analysis, if you backtest enough patterns, you will find that some yield consistently profitable trades (and indeed, it is these patterns that are likely to become popular and be cited by adherents of technical analysis). Thus, it is crucial to test any ostensibly profitable trading rule using out-of-sample data (data different from the one used to identify the rule). When one does this, it often appears that trading rules that are profitable in one time period cease to be profitable in the following time period. Indeed, the authors of the article concluded that technical trading rules ceased to be profitable for U.S. stocks after the late 1980’s. Perhaps this is because their increased use neutralized their efficacy? It’s hard to say. However, Andrew Lo’s description of the stock market as an evolutionary mechanism in which the profitability of various trading rules is always changing appears to hold much truth.

Nevertheless, I am impressed that at least some of the studies covered by this survey article did isolate trading rules that yielded extremely large annualized returns (over 200 percent annualized in some cases). This may suggest that there are some profitable technical trading rules out there that still waiting to be discovered. Of course, the only problem is that once they are discovered and disseminated to the broader public, they may cease to exist. In the meantime, it seems like the only consistent way to make money using technical analysis is to write a newsletter on the subject!

p.s. Thank you to the person who left a comment telling me that the original version of this post actually used a rational number (2/7) and not an irrational one (egg on my face!!!)

Comments:

<< Home

How did you like the sell off today? It is going to get much worse, so just keep holding you clown. You fools will lose it all.

Hey, anonymous, if you don't like the blog, you don't have to here.

If your comments are reflecting your trading, you won't be stay in market for long.

If your comments are reflecting your trading, you won't be stay in market for long.

start nitpick. 2/7 is not an irrational number. It's equal to 0.285714285714(repeating). end nitpick.

BTW I remember seeing OUTL in your portfolio before. It just made a run from 10 to 19 and back. Did you sell it during the runup?

BTW I remember seeing OUTL in your portfolio before. It just made a run from 10 to 19 and back. Did you sell it during the runup?

A few stocks you might find interesting for the new year: NAHC, JBSS, ISH, RML, MHJ, VNT, and DHOM.

Happy Stock Picking,

Sam.

Happy Stock Picking,

Sam.

I've been following DHOM for quite some time. Very cheap relative to book but the company has lots of problems. I used to have 2000 sharse of OUTL but I sold the last bunch of shares at around $11, so I missed the big run up! Sera Sera.

The other stocks mentioned are interesting since they all are trading below book value but I'm not familiar with them. Will need to do more research.

Post a Comment
<< Home