Last issue, we looked at the basic philosophies for stops and stop losses. In part 2 of the series, we will look at some applications of these philosophies as well as historical examples.

Disclaimer: The systems and securities presented here are for demonstration purposes only. Any use of the systems presented is at the sole discretion of the trader. The Trading Systems Analysis Group is not endorsing or denouncing any of the systems shown. No recommendations to buy or sell any security are given or implied. There is a potential for loss in any trading situation.

While the formulas presented are for use with the MetaStock charting analysis platform, the rules can be used with almost any charting software package that allows programming for technical analysis situations. You are given permission and encouraged to experiment with the formulas, to test them, and modify them for your own personal use.

The first step in evaluating any systems’ performance is determining what the true gain is (profit vs. time) in comparison to a buy and hold position. The basis for analyzing a buy and hold position is accomplished with the following system test.

Enter Long: Cum(1)=1

Close Long: Cum(1)=LastValue(Cum(1))

With this system test, we will compare various systems against a buy and hold perspective for one of history’s most popular stocks, IBM. All of the tests on IBM will be done with 1000 periods of data, ranging between April 4, 1996 and March 20, 2000. All tests will be a ‘points only’ test. Rather than starting out with an equity base and counting commissions, we only care how many points a system gains/looses within the approximately 4 year period. In the case of IBM, 82.81 points of profit would have been realized over 1000 periods in a buy and hold position. We will use the term "BAH" to represent the buy and hold results, and all references will be against those values. The BAH equity curve would look like this:

The first system we will look at will be a Linear Regression/Moving Average Crossover System. This system will enter long when the 21-perion Linear Regression indicator (end-point linear regression) crosses above a 21-period Moving Average, and exits when the Linear Regression crosses below the Moving Average. Traditionally (but not always properly) a reversal of a condition on a trend following system infers that the trend itself has reversed. If this is the case, then we will want to exit any position on the opposite condition that it was entered on. Hence, our system would be this –

Enter Long: Cross(LinearReg(C,21),Mov(C,21,S))

Close Long: Cross(Mov(C,21,S),LinearReg(C,21))

And the equity curve of the system (red) compared to the equity curve of BAH (green) would look like this:

In the same period that BAH returned 82.81 points, our system returned only 73.16 points of profit; almost 12% less that the buy and hold method would have produced. Initially, this would seem like an inferior system to the BAH method, but another factor needs to be considered in the equation: the time in the trade. BAH was in a position for 1000 periods, but our system was in a position for only 603 periods. If we were to extrapolate the return of our system to a full 1000 periods of being in a position, then the potential profit would be at over 121 points. This results in a 46% increase in potential to the BAH method. It also results in a safer method than the BAH method.

This would seem that the system is successful. But there can be viewed by some to be a potential flaw in the system. The general idea is that a reversal signal is a change in trend. However, more often than not, trends don’t reverse, they end and prices move into a non-trend state (volatile market). If the trading system is looking for a trend reversal, it may not find one until the middle of a volatile market state. This can result in money being tied-up in a non-profiting position and potentially a loosing one. The concept of using a trailing stop is not to look for the reversal of a trend (the beginning of a new trend) but the ending of the existing one.

To try to meet the needs of a trailing stop approach, the systems' exit will have to be changed for a new philosophy. There is a trailing stop method known as the High-Low Stop. Many people use this as a trading system, but we will see later where it is usually not appropriate as one. The stop is set at the lowest point that trading has occurred over x-number of periods. Most people use a value between 10 and 20 periods, but for stops, we have seen much better results in general with a 5-period value for the stops. The way it works is that you have an entry method to get into a position (in this case, the Linear Regression/Moving Average crossover) and then we will exit our position when the price crosses below the lowest trade over the last 5 days. As prices move higher, the lowest trade over the last 5 days also rises to trail our position. The resulting test will be;

Enter Long: Cross(LinearReg(C,21),Mov(C,21,S))

Close Long: Cross(Ref(LLV(L,5),-1),L)

And the equity curve of the system (red) compared to the equity curve of BAH (green) would look like this:

On first glance, the equity at 75.95 points is only slightly higher that the original system at 73.16 points. But there are two major differences now. The drawdown for this new system has almost disappeared. This is because it is out of the volatile market when a trend doesn’t exist. This is a mixed blessing though, since you can also be prematurely taken out of a trend without an appropriate signal to get back into a relevant trend (this occurred in the system shown). The other major difference is how many trading days were used to acquire the profits. The new system was only in a position for only 341 periods as opposed the BAH periods of 1000. This new information extrapolates to an over 222 point gain for 1000 trading periods (a 168% increase over BAH alone).

As with entry strategies, there are many (if not unlimited) exit strategy/trailing stop methods that exist and are functional. A problem is trying to determine what kind of exit method works best for your particular entry method. Different strategies have different speeds and faster ones will cause whipsawing that result in either losses or giving up potential profit. Others don’t trail fast enough and typically result in losses or giving up profits already acquired. It is very important to narrow an exit strategy to the entry method.

An example of another system follows. This is based upon a combination of Welles Wilders’ Directional Movement indicators and Parabolic SAR indicator. The basis for the first system is a traditional usage of many market experts. It involves entering when the 18-period ADX is rising and the Plus Directional Movement (PDI) is greater than the Minus Directional Movement (MDI). It exits when either the ADX is falling or the PDI crosses below the MDI. The system would be this –

Enter Long: ADX(18)>Ref(ADX(18),-1) AND PDI(18)>MDI(18)

Close Long: ADX(18) And the equity curve of the system (red) compared to the equity curve of BAH (green) would look like this: