I’ve spent the last 2 days modeling and testing a trend-following system that I adapted from Curtis Faith’s Way of the Turtle. It’s a trend following system since it’s based on a moving average, but the twist is that it utilizes a moving average band instead of a line. This band helps to weed out some of the whipsaws that kill a traditional moving average system (which, as I’ve discussed earlier, is well and truly dead). Hence, the system only goes long or short if and only if the closing price crosses above or below the band respectively.
The band is created by adding or subtracting a multiple of the Average True Range (ATR) from the moving average. I also varied the multiple – for example I could allow for a larger band above the moving average and a smaller band below it. It seems to work very well, especially for markets that experience slightly bigger pullbacks in the bullish or bearish direction.
So far, results have been promising: I tested it out on the Dow Jones mini contract, SGD/USD, TNotes and Gold. With the exception of Gold, the system performed admirably in the other markets. I evaluated “perform well” by how it did in out-of-sample data as well as in-sample data. Most tests yielded Net Profit/Max Drawdown ratios of 5 or above for both in-sample and out-of-sample data, which means that it is probably robust.
The system didn’t perform well in Gold because of a pecularity of the market. I have about 12 years of data, but gold was marked by a steady, low-volatile rise for the majority of the time period. It only started getting extremely volatile in 2007, hence most of my optimization yielded parameters that were suited for bullish, low-volatility markets, which I doubt will be the case from now on. I think gold will be a very tricky market to test and trade.
However, I am very excited by the results of the other tests, especially for the currencies markets. Currencies are traded in lots of 10,000, hence profits and losses are only in the couple of hundreds, which allows me to trade more contracts and more accurately vary the contract size according to risk.
This looks like a promising system to pursue.