A Test To Find The Best Moving Average Sell Strategy: Comments and Observations
By Dr. Winton Felt    

These comments are in reference to Our Previous Study:   A Test To Find The Best Moving Average Sell Strategy     

Some have recently raised a question about the near absence of tests using exponential moving average systems.  We once performed a study in which we compared dual exponential and simple moving average systems.  The actual data on that is unavailable right now, as explained below.  However, the simple moving averages were generally superior in all ranges tested.  That, and the size of the project, is why we focused more on simple moving averages in the large test.

We also tested systems in which the signal is generated by a closing price cross of the moving average.  For the latter, we tested all moving averages between 4 days and 200 days.  The tests covered 426 stocks over at least 9 years.  Though these tests did not cover thousands of stocks as in the above study, the results were consistent enough that we considered them to be meaningful.  Again, the simple moving averages were generally superior.

We did some limited testing in which we compared SMA and EMA systems as complete strategies.  That is, we both bought and sold using those strategies.  Far more times than not, the simple moving average systems outperformed the exponential systems.  The exponential moving average is a more sophisticated approach, and it does react more quickly to recent price behavior.  However, the simple moving average systems generated greater bottom-line profits.  That surprised us.  It was counter-intuitive.  Our conclusion from this was that the relative slowness of the simple moving average systems allowed momentum to build a little more (and allowed the trend to gain a little more "steam") before the buy signal was generated.  This probably resulted in fewer "false starts" and therefore reduced whipsaws for most stocks most of the time, though we have not verified this.  What about the reported experience of some traders that simple moving averages do not handle sharp price swings as well as exponentials, resulting in the short moving average whipsawing across the longer moving average?  It can be argued that while it might be a handicap some of the time, it is apparently not a handicap most of the time, inasmuch as the simple moving averages produced better bottom line results most of the time with most stocks tested (425 stocks).  It may be that better entries (entering when trends were more mature) more than compensated for any extra whipsaws (fewer "bad starts" may have reduced the number of "bad exits").  For whatever reason, the frequency with which it was a problem was not enough to shift the balance in favor of EMAs for most situations, though EMAs were superior in some instances.  A note of clarification should be made here.  There were many instances where the EMAs outperformed the SMAs, and if a person were to use an optimization program to find the best system for trading a particular stock, that system may very well turn out to be an EMA system.  However, our purpose was to find the approach that worked best for most stocks most of the time.  That is why we did not perform the tests on a short list of 50 stocks over a period of only two or three years.

The notion that SMA systems are more susceptable to whipsaws might be a more relevant consideration if an individual is trading particularly volatile stocks or commodities.  However, the link below will take you to the conclusions of another study that addresses that very issue with respect to commodities.  We have said that some dual EMA systems outperformed equivalent SMA systems.  In many cases we found that simply using a slightly different combination of SMAs improved results to a level that could not be matched by making similar adjustments to the EMA system.  This was not always so, but it does suggests that the problems associated with SMA systems in a volatile environment can often be compensated for by simply adjusting the lengths of the averages used.  Many traders prefer to use the EMA over the SMA, primarily because of its sensitivity to the most recent price action.  The perception of those traders is that the use of a faster more responsive system is an advantage.  Our intuition on the matter is that this is obviously true.  However, the fact remains that, despite our intuition on the subject, speed and responsiveness do not necessarily correlate with profitability.

The tests in the above report were not the only dual moving average tests we conducted.  We covered all dual systems in which the shorter moving average was between 4 days and 50 days and the longer moving average was between the short moving average in length and 200 days.   In posting only the above information, we were attempting to narrow down a lot of data to the few system variations we perceived to be of interest to most people.  We have highlighted in bold blue fonts the systems that were of particular interest at the time.  After we culled that information from the larger study, the rest of the data from the study was misplaced (they are undoubtedly somewhere in one of our storage systems under an incorrect or non-intuitive title.  A recent preliminary search for that data was unsuccessful.  We may eventually publish it on this site, if we can find it.

Despite the conclusions of the last paragraph of the above report, there was really very little difference in bottom-line performance between the top-ranked 10-20 system (return = $8,162,053) and the eighth-ranked 9-18 system (return = $7,964,701).  Remember that the performance figures were generated by using all the systems on thousands of stocks over many years and summing the profits accumulated by each system.  This study was not designed to discover which system performs best as a complete trading system.  Because the 9-18 system will generally exit a position faster than the 10-20 system, the trader using it might conceivably enter a new position before and at a better price than the trader using the 10-20 system.  On the other hand, the slower entry may give the advantage to the 10-20 system.  The real difference in performance is more likely to be execution rather than the particular system employed.  The individual who is highly disciplined in implementing the 9-18 system, is likely to do better than a trader who is not quite as disciplined while implementing the 10-20 system.  

We now wish to add one more wrinkle to the idea expressed in the previous paragraph.  If you prefer EMA over SMA systems, then use them.  You are more likely to be disciplined in the execution of a system if you believe in it.  Again, it is not the particular system you use as much as it is the discipline with which you implement the system that will determine profitability (assuming you are using a system of a quality similar to one of those in the top eight of our study ... whether SMA or EMA).  Traders often forget that they themselves are part of the system they are implementing.  The bottom line profitability of any trader will depend on how that trader and his or her system "dance" together.  If you have inner conflicts or doubts about your system, those conflicts and doubts will interfere with the execution of your discipline, and they will therefore impact profitability.

For additional information on a comparison of simple and exponential moving averages, we refer the reader to the following link.  It will take you to a study conducted by Merrill Lynch.  We do not have a copy of the original study, but the link will take you to a quote from John J. Murphy, Technical Analysis of the Futures Markets: A Comprehensive Guide to Trading Methods and Applications (New York,: New York Institute of Finance, 1986), p. 252.  This quote includes four conclusions made by the researchers.  A Merrill Lynch Study Comparing Moving Average Trading Systems