Hi Don,
Nice work! You are always so thorough.
I have one disagreement and 2 suggestions. I disagree that one can tune the moving average to a particular stock. Stocks don't have character in my opinion. It is all market psychology. Having said that, a way to prove that tuning works would be to go back into history with a cut-off date, for example 1980. In 1980, find the best moving averages that worked for a stock up to that point. Use those moving averages for 1981. At the end of 1981, retest the best moving averages for your stock. Use those moving averages for 1982. Repeat this process. I'm fairly certain you will find no advantage to tuning.
Suggestion 1: Run the same MACRO test using the MattMod safe.
Suggestion 2: I have been wanting to run a universal test using Stochastics (or some variation). Moving averages lag, but stochastics anticipates. The problem with Stochastics and its ilk is that there are many anticipations of "not much." However, this is where a flavor of AIM would beneficial - preventing action on the "not much" signals.
If you ran stochastics on weekly data at a setting that equaled 6 months and only took signals in the upper/lower zone at the first down/up tick, I think you would find that your buys and sells would be at better prices than moving averages.
I would be interested in just such a study. I have not had a chance because I have been consumed with finishing a large project in another investment vehicle. So, if you get a "hankering" to do it, I for one would be greatly interested.
Anyway, you've done fine work. I and others appreciate it.
Your friend,
Matt