 |
This month's issue of STOCKS & COMMODITIES offers a first: a solution to the perennial question of determining the optimum filter length (in practical terms, moving average length) to use when analyzing pricing. If we knew that on a daily basis, all our indicators could be optimized for what's actually happening in the market. How, you ask? Well, most indicators require input as to the length of data to analyze in their computation. Up to now, that length has always been a best guess, but our lead article this time tells us how we'll now have an objective number from which to work.
In "Adaptive Trends And Oscillators," S&C contributor John Ehlers, expanding on his article in the March issue, has not only developed a specific analysis but published the code for it, both here and in the March issue. The technique is definitely high-tech, but the code is usable by anyone with EasyLanguage access. Programmers will have no trouble converting it to any other language.
Not only that, Ehlers throws in an example by creating an oscillator with a unique property as a result of his solution: it rarely gives bad signals in a trend. Anyone with any experience using technical analysis' plethora of oscillators knows they're great when prices aren't moving too far or are actually oscillating around a relatively narrow range of prices. But when a firm move begins, any reaction to the move generates a reaction trading signal that's quickly wiped out when the trend resumes. Ehlers's oscillator beats this and thus bids to become the ultimate oscillator, a single indicator you can use in both trending and ranging phases.
Has Ehlers found the Holy Grail of trading? Close, but not quite! Because the amount of cyclic content in a price series varies constantly, this indicator, like any other, can be thrown off. If cyclic content is weak, the indicator will be weak. Nor can it be tweaked by playing with the parameters; it's a solution, not a formula whose performance can be adjusted until it works properly. There are no parameters as Ehlers publishes it, though I'm sure techies will want to insert some.
How does the new solution compare to other estimators of filter length? Ehlers feels it's almost as good as maximum entropy spectral analysis (MESA), but I think it's better because MESA is always a best estimate, not a precise solution. Practically speaking, it's ahead of Fourier transforms, estimating averages, or playing with cycle wands on a screen.
Ehlers has already converted this work to commercial trading systems, but he's published his work here for S&C readers. The trading world is full of secrets that don't need to be, and perhaps his example will inspire others to put their work in a public forum (ours, of course), where it can only be improved by the efforts of other minds.
This article is definitely worth checking out. Remember, you saw it here first!
Good Fortune!
