By Doug Tucker
Most indicators that come pre-programmed in technical analysis charting software use static input parameters. This is the simplest solution, but not necessarily the best.
Many indicators were originally designed for use on daily data, so an average cycle length was estimated based on a monthly cycle. Many momentum indicators by default had an input parameter, or lookback period, of 14 bars, supposedly reflecting a half cycle of 28 days in the month, since most momentum indicators are best using one half cycle period. However, there are only about 21 trading days per month, and a cycle measure of 20 is more accurate than 28 as an average on most daily charts. In my opinion the 14 period parameter just stuck because so many people started using it, even though it was probably based on a false assumption. Some charting software, by default, used a 9 period lookback rather than 14, probably in an attempt to reflect a more accurate monthly cycle.
Indicators that are moving average based should probably use a full cycle length, as the purpose of a moving average should be to filter out the noise in the market, and then leave the underlying trend. If noise isn't first filtered out, then the indicator is just giving a signal based on noise, rather than the filtered output, which should be the trend you want to track. A common problem with the CCI, or Commodity Channel Index, is that it commonly uses too short an input parameter. Lambert, the originator, originally thought 1/3 of a cycle would be good for the lookback period, but that didn't test out very well. Many trading chat rooms use a 14 period input, but that is far too short, and gives signal based purely on noise. Of course, if the cycle is so short that the 14 parameter is correct and reflecting the current cycle, you probably shouldn't be trading at that time because the market is most likely in chop.
The problem is in choosing a parameter, if you accept the above premise, is that the cycles keep changing. If markets traded in nice, even cycles, it would be a simple matter to select the correct input for the type of indicator you are using.
There have been many attempts to make moving averages and other indicators adaptive to volatility or some other characteristic, but rarely to make them adaptive to the actual cycle length. Most early attempt to measure cycles were visual estimation based on counting swing lows. This is fine to get a ballpark estimation, but it is difficult to keep changing the parameter on the indicator every time the cycle length shifts. And, it makes it difficult to test a strategy. It is better to let the computer estimate the cycle length.
The common way to measure cycles, if the cycles are nice and even such as a sine wave, is to use a Fast Fourier transform. But the problem with this approach is it takes many cycles before the output of the transform is stable and reliable. The better approach is to use a method that can try to extract the cycle as it is forming; using a method referred to as "instantaneous frequency measurement."
John Ehlers has done much excellent work in this area. There is no way I can do justice to his work in the few hundred words here, but there is much written on the subject on the internet and in print. I have a link to his website on my website (see link below) in the resources tab, as well as references to a couple of his books on my book tab. Many of his formulas are written out on either his web site or in his books, and probably elsewhere on the internet
Also, I have an article in the indicator tab on my web site called "CCI- Making it Better" that has many charts and examples comparing using static input parameters and using adaptive parameters. Although the focus of that article is the CCI, the same principles can be applied to a stochastic or the RSI, and most other indicators. Keep in mind you need to use just half the cycle measure on many of the momentum type indicators.
Once the type of cycle measure is decided, it is a simple matter to re-write the indicator in you charting software programming language, and then use the output of the cycle measure code in place of the static number. By making the effort to do this, an average indicator can become much better when it knows what cycle it is trying to measure and track.
This article is not meant to be a complete tutorial on the subject of using adaptive techniques to improve and enhance indicator usefulness. The purpose is just to be an introduction on the subject. If you are interested, there are many free resources on the internet to explore this topic in more detail. It is well worth the effort. When using technical indicators it is best, in my opinion, to find an approach not everyone else is using. One needs to find their own way, and stay out of the crowd.
Doug Tucker has a blog with daily commentary on stock indexes, precious metals, and other markets. There are many articles on technical analysis and indicator design and interpretation. To visit go to: http://tuckerreport.com/
Article Source: http://EzineArticles.com/?expert=Doug_Tucker
http://EzineArticles.com/?Using-Adaptive-Parameters-in-Technical-Indicators&id=672282
No comments:
Post a Comment