The Hull MA and Regression Lines

Ensign Software recently introduced Alan Hull's moving average formula as a downloadable template. Mr. Hull, who kindly provided an article for the Ensign Newsletter wherein he discusses the formulation and compares it to standard moving averages.

Being curious, I naturally tried it and indeed found that it tracks price action quite nicely. Being an engineer, I then proceeded to compare it to other methods I already understand, in this case, a Regression curve, which is the center line of the Regression Channel study. The Regression channels themselves are optional and are built by specifying a fixed percentage above and below the center line.

What I discovered was very interesting to me. I found that the Hull MA and a comparable period Regression curve are extremely close at small moving averages, eg, 26-bar and smaller. Above this value, the divergence between the two is more noticeable. Here are three charts that illustrate this.

The next decision was whether to use it in my trading. In my case, the central approach of my trading is "actio-et-reactio", action begets reaction, price seeks balance by moving from one extreme through a balance point to continue to the other extreme. Regression curves embody exactly this principle and so they are a natural tool for my way of thinking.

Regression curves sample the last n-bars of data using a "least squares" method to produce a best-fit line of that data. As a new bar is made, the fist bar in the series is dropped and the best-fit calculation is done again using the new point. Howard Arrington wrote an article about this method, which you can read here.

Having used regression methods now for several years, I've learned to spot interesting price behavior at regression curve inflection points. Not fully understanding Mr. Hull's implementation, nor having the experience using it, I wasn't able to explain to myself the divergence I saw between the two at higher values. For this reason I decided to stick with regression, which I do understand. This in no way says anything about the quality of the Hull MA, which may indeed be preferable should I study it in greater depth.

My conclusion is that the use of the Hull MA is largely a matter of preference, and for those who seek faster tracking at inflections, is preferable to a simple, exponential, or weighted MA.

Regression Curves & the Hull MA