By definition, trend-following systems never sell near the high or buy near the low, because a meaningful opposite price move is required to signal a trade. Thus, in using this type of system, the trader will always miss the first part of a price move and may surrender a significant portion of profits before an opposite signal is received (assuming the system is always in the market).
There is a basic trade-off involved in the choice of the sensitivity, or speed, of a trend-following system. A sensitive system, which responds quickly to signs of a trend reversal, will tend to maximize profits on valid signals, but it will also generate far more false signals.
A nonsensitive, or slow, system will reflect the reverse set of characteristics. Many traders become obsessed with trying to catch every market wiggle. Such a predilection leads them toward faster and faster trend-following systems. Although in some markets fast systems consistently outperform slow systems, in most markets the reverse is true, as the minimization of losing trades and commission costs in slow systems more than offsets the reduced profits in the good trades.
This observation is only intended as a cautionary note against the natural tendency toward seeking out more sensitive systems. However, in all cases, the choice between fast and slow systems must be determined on the basis of empirical observation and the trader’s subjective preferences. There is a wide variety of possible approaches in constructing a trend-following system. In this chapter we focus on two of the most basic methods: moving average systems and breakout systems.
Moving Average Systems
The moving average for a given day is equal to the average of that day’s closing price and the closing prices on the preceding N − 1 days, where N is equal to the number of days in the moving average.
For example, in a 10-day moving average, the appropriate value for a given day would be the average of the 10 closing prices culminating with that day. The term moving average refers to the fact that the set of numbers being averaged is continuously moving through time.
Because the moving average is based on past prices, in a rising market the moving average will be below the price, while in a declining market the moving average will be above the price. Thus, when a price trend reverses from up to down, prices must cross the moving average from above. Similarly, when the trend reverses from down to up, prices must cross the moving average from below.
In the most basic type of moving average system, these crossover points are viewed as trade signals: a buy signal is indicated when prices cross the moving average from below; a sell signal is indicated when prices cross the moving average from above. The crossover should be determined based on closing prices.
in my view, there is no strong empirical evidence to support the idea that linearly or exponentially weighted moving averages provide a substantive and consistent improvement over simple moving averages. Sometimes weighted moving averages will do better; sometimes simple moving averages will do better.
The question of which method will yield better results will be entirely dependent on the markets and time periods selected, with no reason to assume that past relative superiority will be indicative of the probable future pattern. in short, experimentation with diff erent weighted moving averages probably does not represent a particularly fruitful path for trying to improve the simple moving average system.
Generally speaking, the crossover moving average system is far superior to the simple moving average. (However, it should be noted that by including some of the trend-following-system modifi cations discussed in a later section, even the simple moving average system can provide the core for a viable trading approach.) The weaknesses of the crossover moving average system and possible improvements are discussed later.
Breakout Systems
The basic concept underlying breakout systems is very simple: the ability of a market to move to a new high or low indicates the potential for a continued trend in the direction of the breakout. The following set of rules provides an example of a simple breakout system:
- Cover short and go long if today’s close exceeds the prior N -day high.
- Cover long and go short if today’s close is below the prior N -day low. The value chosen for N will defi ne the sensitivity of the system. if a short-duration period is used for comparison to the current price (e.g., N = 7), the system will indicate trend reversals fairly quickly, but will also generate many false signals. in contrast, the choice of a longer duration period (e.g., N = 40) will reduce false signals, but at the cost of slower entry.
The following three observations are also valid as generalizations describing the trade-off s between fast and slow breakout systems:
1. A fast system will provide an earlier signal of a major trend transition
- A fast system will generate far more false signals.
- The loss per trade in the slower system will be greater than the loss for the corresponding trade in the faster system. In some cases, a fast system might even realize a small profit on a minor trend that results in a loss in a slower system. For example, the N = 40 system’s August buy signal that was liquidated in November resulted in a net loss of approximately $2.54 (excluding commissions). The corresponding buy signal for the N = 7 version—triggered in July and exited in September—resulted in a net gain of around $2.46.
As indicated by the preceding illustration, fast and slow systems will each work better under different circumstances. In the case of the chosen illustration, on balance, the slow system was much more successful. Of course, one could just as easily have chosen an example in which the reverse observation was true. However, empirical evidence suggests that, in most markets, slower systems tend to work better. In any case, the choice between a fast and a slow system must be based on upto-date empirical testing.
The previous example of a breakout system was based on the current day’s close and prior period’s high and low. It should be noted that these choices were arbitrary. Other alternative combinations might include current day’s high or low versus prior period’s high or low; current day’s close versus prior period’s high close or low close; and current day’s high or low versus prior period’s high close or low close. Although the choice of the condition that defines a breakout will affect the results, the differences between the variations just given (for the same value of N) will be largely random and not overwhelming. Thus, while each of these definitions might be tested, it probably makes more sense to focus research efforts on more meaningful modifications of the basic system. The pitfalls of breakout-type systems are basically the same as those of moving average systems and are detailed in the following section.