System design, part 11: A wider perspective

By Active Trader Staff
In researching and testing our pullback system, one perceived weakness has been the relative infrequency of trades, which became especially evident when we moved out of an initial 10-year test period (1994-2003) and applied the system to new data. A version of the long-entry rule that automatically adjusted for volatility (continuing to trigger trades when the original setup lay dormant) showed potential for addressing that problem on a limited basis, but hasn’t proven it offers a significant advantage over the long term.

We’ll experiment with a different idea here: Adding a second pullback pattern, equivalent in principle to our existing long-entry setup, but different in its specifics. After analyzing its performance, we’ll apply our system rules to different markets. For the new long-entry pattern, a buy is signaled when:

1. Today’s low is less than yesterday’s low.
2. Yesterday’s low is less than the previous day’s low.
3. The previous day’s low is less than the low the day before that.
4. Today’s close is less than today’s open.
5. Today’s close is less than the close two days ago.
6. Yesterday’s close is less than the close three days ago.

In formula form these rules are:

1.    low[0] < low[1]
2.    low[1] < low[2]
3.    low[2] < low[3]
4.    close < open
5.    close < close[2]
6.    close[1] < close[3]

where 0, 1, 2 represent today, one day ago, two days ago, etc.

Unlike the original entry pattern, this setup is much more robust in that it has no optimizable parameters, such as a defined price move from one bar to the next; it’s simply the relative price levels (higher or lower) of the low, open, and closing prices over the four most recent days. Figure 1 shows examples of the new pattern along with the volatility-adjusted version of the system’s original long-entry signal. At first glance, anyway, a benefit of the new signal is that it captures trade opportunities the original signal misses.



Table 1 compares the results of trading the two patterns along with the original short setup, using the test conditions outlined in “System rules.”



Both versions of the system were partnered with the momentum-indicator exit for long trades and a five-day exit for short trades; all trades were also offset by entry signals in the opposite direction. In most categories, the results for the Entry 2 version are comparable, but slightly worse, than the original version’s results, but Entry 2 produced 183 trades vs. only 149 for the original long-entry version. However, these additional trades simply translated into a 5.31-percent decrease in total net profit for all trades. The long-side performance was very similar for both versions, but Entry 2 apparently didn’t complement the short-side signals as well — profit declined nearly 16 percent in that category, from $6,001 to $5,060.

For the complete article, see the April 2010 issue of Active Trader magazine. Click here to subscribe.



|

|