The biggest problem with a basic moving average crossover system is that it is always in the market. This leads to many false signals and whipsaws. It also tends to give back a large portion of its profit as the slower moving average lags the faster one.
Looking for a way to improve on this system, I came across the Triple Moving Average Crossover System. This system operates in a similar manner, but adds a third moving average to confirm entries and create earlier exit signals. In theory, this will create less false entry signals and retain more profits on successful trades.
About The System
This system uses three different moving averages to asses the direction of a trend and then follow it, eventually exiting the trade when the trend ends.
An entry signal is generated when the fastest moving average crosses the slowest moving average and then crosses the middle moving average. The trade is then held until the fastest moving average crosses or touches the middle moving average. The system also sets an initial stop at the high or low of the most recent price swing.
One of the most interesting aspects of this system is that you can create an almost unlimited number of variations based on the moving averages you elect to use. The most common approach is to use Exponential Moving Averages (EMAs) for shorter term systems, and Simple Moving Averages (SMAs) for longer term trading ranges
For shorter time frames that trade one-hour bars or faster, using 4 EMA, 9 EMA, 18 EMA or 10 EMA, 25 EMA, 50 EMA is recommended.
For longer time frames that trade daily or weekly bars, using 4 SMA, 10 SMA, 50 SMA or 5 SMA, 10 SMA, 20 SMA is recommended.
For the following backtesting results, these are the parameters used:
Market: Great Britain Pound vs Japanese Yen
Period: 1 Hour Bars (Jan 14, 2005 through October 12, 2011)
Fast Moving Average: 10 EMA
Medium Moving Average: 25 EMA
Slow Moving Average: 50 EMA
Go Long When:
- 10 EMA crosses above the 25 EMA and then the 50 EMA
Go Short When:
- 10 EMA crosses below the 25 EMA and then the 50 EMA
Exit Long When:
- 10 EMA crosses below or touches the 25 EMA
Exit Short When:
- 10 EMA crosses above or touches the 25 EMA
An investment of $10,000 in this system would have returned a profit of $6,958.64 in the period tested. This represents a return of 69.6% in six years and ten months. It is very interesting to note that the S&P 500 is just slightly up over the same time period.
The system posted a profit factor of 1.10 and and expected payoff of 6.26. It had a maximal drawdown of 22.79% and a relative drawdown of 26.05%.
It’s largest profit was 3216.57 and its average profit was 230.17, while its largest loss was 729.36 and its average loss was 95.86. The system was profitable on 31.32% of its trades.
This system has a lower Win Rate and a lower Payoff Rate than the SPY 10/100 Long Only System, but it also trades more than 27 times more frequently. The vast number of trades allows it to make up for its lower Win and Payoff Rates and actually become more profitable.
The goal of adding the middle moving average was to improve the system’s ability to hold on to profits. This likely contributed to the Triple Moving Average Crossover System having a far lower drawdown than the SPY 10/100 Long Only System.
Despite having a lower Win Ratio and Profit Ratio, the fact that this system can make profitable trades as frequently as it does with such low drawdowns makes it absolutely worth further review and testing.
My main reservation about these test results are that they are exclusive to one market over one seven year period. I would be very curious to see if the system can generate similar results when tested across multiple different markets and time periods. This would prove that the system is not just benefiting from testing an ideal scenario.
It would also be interesting to test this system using a wide range of moving averages. While the results we have prove that the system is worth developing, I am curious whether adjusting the moving averages could either improve or ruin the system’s returns. There are endless combinations that we could test, but I would be especially interesting in how adjusting the middle moving average affects drawdowns.
One of the key aspects of moving average systems is that they perform best in trending markets and generally underperform in range-bound markets. Adding a component that would allow you to distinguish between trending and range-bound markets and only trade the system in the trending markets might be very profitable. It also, however, might result in curve-fitting that could backfire down the road.