One of the first systems I wrote about here was the Simple MACD / 200 SMA System. I have always thought that the MACD indicator was an incredibly useful evolution of moving averages. However, I recently found testing results that proved the exact opposite.
While the results were not what I expected to find, they did provide a fascinating lesson on how to handle results that are contrary to expectations.
MACD Testing
In March of this year, Derry Brown from System Trader Success published an article discussing the results of his extensive testing of the MACD indicator. He compared 250 different combinations of fast EMAs, slow EMAs, and signal lines. The fast EMAs ranged from 10 to 50, the slow EMAs ranged from 2x to 6x, and the signal line ranged from 2 to 20 in multiples of 2.
Using historical market data, the article assumed a long position when the MACD Line was above zero and above the signal line. The results were displayed were adjusted to represent annual returns with respect to the amount of exposure. The actual returns numbers are insignificant. What is interesting, is the differences between the numbers depending on the variables.
Testing Results
For almost every set of EMAs, trading the MACD produced better returns when the signal line variable was the smallest. The article further concluded that the best performing variable combination included a fast EMA of 1. Of course, an EMA of one would be equal to the price itself, so the best performing MACD variables are actually measuring the convergence and divergence between the price and a moving average.
This led Brown to conclude that the MACD indicator is less accurate than using standard moving averages when it comes to long term trading. It is more effective for ultra short-term trading, but its edge in those situations is erased by high transaction costs. The only place that the MACD did seem to excel was on mid-term short side trades using a fast EMA of 16, a slow EMA of 97, and a signal line of 2.
Applying These Results
The knee-jerk reaction to these testing results would be to abandon the MACD’s use in all situations except for medium-term short side trades. That may not be the best idea, though.
Underperforming another indicator by no means makes the MACD useless. It may still be very useful when partnered with some other indicator. Perhaps teaming the MACD up with the 200 unit SMA like my first system post suggested would diminish the negatives pointed out in these test results. It may also have potential as a confirmation indicator.
We have to remember to give enough credit, but not too much credit, to one individual test. While the MACD may not be the optimal indicator to use by itself, it may still have plenty of value in other situations.
The Deeper Lesson
At the end of the article, Brown actually states that he was a strong supporter of the MACD indicator and that he fully expected it to outperform moving averages across the board. This is what I found to be the most educational part of the entire article.
Like the author, we have to be willing to accept backtesting results that provide evidence contrary to our beliefs. One of the keys to successful system building is taking an honest, unbiased look at backtesting results and being willing to scrap something that isn’t working, even if it seemed like a brilliant idea.
This can be the hardest part of system building, because it messes with our ego. Of course, if we want to be successful in this business, that ego needs to be checked at the door. Setting your ego aside in order to optimize your system is the mark of a successful systems trader.