All About Moving Averages
There is probably more actual money being traded today using moving averages than with all other technical indicators combined. Because they can be used for everything from finding very long-term monthly trends to setting stops for day trading, and for anything in between, moving averages have been the subject of more discussion in technical literature and elsewhere than any other study. One reason they enjoy such popularity is when the markets trend, these simple little lines work as well or better than indicators, which require a Ph.D. to interpret.
Moving averages smooth out market fluctuations and short-term volatility and give the trader some indication on which way financial market is going. Just as important is what they don't do. Unless you plot them as an oscillator, they provide no overbought/oversold information at all.
They are trend-following indicators of the purest sense. They show the direction of a trend, yet they don't measure how strong or weak the trend is. Their function is to identify the direction of the trend and then smooth out or dampen its volatility. Moving averages do these things very simply and very well.
Types of Moving Averages
There are so many possible types and combinations of moving averages that it is impossible to list them all. Most of the more exotic moving average varieties were created in the 1970's, when moving averages were considered to be very sophisticated and advanced technical analysis tools.
Today, a lot of talented and creative virtual trader market technicians spend most of their time figuring out new ways to improve and trade moving averages. This interest in moving averages was well rewarded in the past, during times of endlessly trending markets, when moving averages worked extremely well. Most of the more inventive types have since fallen into disuse (like Maxwell's "modified accumulative" or "average-modified"). Three major categories have survived the test of time: simple, weighted and exponential.
Simple Moving Averages
The simple moving average is calculated by adding and then averaging a set of numbers (usually daily closes) representing market action over some specified period of time. The oldest data point is dropped as a new one appears; thus the average 'moves' and follows the market. A line connecting the daily averages will have the effect of smoothing recent market volatility. A large data set representing a large amount of past data will create a smooth line. A smaller data set representing only more recent data will create a more responsive line.
Weighted Moving Averages
The simple moving average gives equal weight to each point in the set of back data. The weighted moving average assigns greater importance to more recent data by weighting each day's data differently. This is usually done by multiplying the most recent data point by a given number, adding the result to the overall calculation, then multiplying the next most recent point by a lesser number, and so on. The resulting line will be more responsive to recent market activity than the simple moving average.
Exponential Moving Averages
The simple and weighted moving averages are only capable of reflecting the data in the number of data points chosen for the calculation. The exponential moving average assigns greater importance to more recent market action like the weighted moving average. It also takes into account all of the data in the data set, leaving nothing out. A daily exponential moving average uses the life of a futures contract. A weekly or monthly exponential moving average uses as much data as you give it.
Despite the seeming sophistication of weighted and exponential moving averages, nearly every test we've seen or done ourselves has shown the simple moving average to be superior to the others in terms of trading results. Our own research indicates that weighting the data to emphasize recent events makes the indicator overly sensitive, negating the original purpose of the moving average: to smooth out market action.
Weighted and exponential moving averages tend to generate more trades in tight, trading-range markets than simple moving averages. The result is usually costly whipsaws. We recommend using simple moving averages only. Save your system's complexity for other things, like money management and risk control.
Multiple Moving Averages
Moving average based futures market or stock market daytrading systems can use either a single moving average or any number of moving averages in various combinations. We've used single, dual, triple, or even quadruple moving average systems. We suppose that any multiple is possible, but the variations with only three or four can easily become overly complex and frankly, we see no advantage to using anything more complicated than necessary.
Single Moving Average Systems
The simplest and often the most effective moving average is the single moving average. It is most useful as a long-term trend indicator, rather than as a daily trading device. For example Colby and Meyers, in their book The Encyclopedia of Technical Market Indicators, optimized for a single moving average over 75-years of monthly NYSE data, using a simple reversal system. They found 12-months to be the optimum number, beating a buy-and-hold strategy by a large margin. In our experience, this simple 12-month moving average is a stock market timing device that's hard to beat.
The basic rules for trading with a single moving average are simple: buy when prices rise above the average, sell when prices drop below the average. This results in a simple reversal system that is always in the market. As you might imagine, a reversal system is very susceptible to whipsaws and should only be used as a long-term trend indicator.
Dual Moving Averages
The most popular moving average systems use two moving averages. These generally consist of a longer-term average that serves to define the trend, and a shorter-term average that gives trading signals as it crosses the longer-term average. The best known of these is Richard Donchian's 5-day vs. 20-day system.
Most research we've seen shows that dual moving average systems tend to be more profitable than other types. The research also shows that they, like just about all moving average systems, have their ups and downs. They are notorious for giving back too much of their hard earned profits. Anyone who traded Donchian's system (which, by the way, is not a simple reversal system, but uses an elaborate set of filters) during the trending 1970's made regular and substantial profits. The same system lost heavily during the middle and late 1980's. The basic signal with two moving averages is the crossover. Buy when the shorter average crosses over the longer, and sell when the opposite occurs.
Three Moving Averages
The most popular triple moving average is the widely followed 4-9-18-day method popularized by R. C. Allen in the early 1970's. The third moving average opens up a huge number of potential trading possibilities.
Generally speaking, when a market has bottomed the major indication of a trend change is the 4-day crossing the 18-day. The confirming signal is the 9-day crossing the 18-day. As prices peak, the preliminary indication of a possible trend change will be the 4-day crossing the 9-day. This may be an early point to take profits. The trend reversal will be completed only when the 4-day and the 9-day cross the 18-day.
We like the triple moving average systems because they offer the advantage of a neutral zone as opposed to the reversal trading called for by the single or double moving average methods.
For example, in the 4-9-18 system discussed above, when the 4 crosses the 9 we exit our position and we don't take a new position in the market until the 9 crosses the 18.
We also like the triple system because the 4 crossing the 9 is a quick profit taking mechanism that overcomes some of the problem of giving back too much profit that we mentioned before. We believe that exits should always be quicker than entries in a successful trading system.
Four Moving Averages - Overkill?
Using four moving averages is neither as strange nor as difficult as it sounds. Used properly, the four moving average approach addresses some of the problems inherent in moving averages while losing none of the advantages. The method uses the four moving averages in sets of two. The two longest moving averages are used strictly as trend identifiers and are most easily utilized when set up as an oscillator moving about on either side of a zero line. The two shorter moving averages are more sensitive and are used for timing the entries and exits (usually on a crossover basis) only in the direction of the longer-term dual oscillator.
Trades against the trend are, by definition, filtered out. If an uptrend exists (as defined by the longer-term oscillator) only long trades will be taken as signaled by the shorter-term crossovers. Conversely, only short trades will be taken in a downtrend. There will be neutral periods during trend corrections. There will also be neutral periods during sideways markets. Whipsaws will not be eliminated, but they will be significantly decreased.
Most traders use a variety of filters to decide if the initial signal is valid. Filters come in two categories: price filters and time filters.
1. Filtering signals by price normally means waiting for the price to meet some additional criteria before entering the trade. This might be determined by measuring the amount the price has penetrated the moving average or measuring the distance that one moving average has crossed over another. The trader in this instance is looking for confirmation that the moving average crossing was not a random price event but is indeed a trend change. The new trade is not taken until the price has exceeded the moving average by a minimum amount. Another variation of this filter would be waiting for prices to move by a given percentage after a crossover. Our favorite filter or confirmation method is simple: wait until you get a close in the new trend direction.
2. Time filters involve waiting a number of time periods after the crossover before trading in the new direction. Many moving average traders have observed that most of the whipsaws are very immediate and a slight delay in entry can avoid most of them. The waiting period would typically be from one to five-days. If the price stays on the new side of the moving average for the minimum time period, it is assumed that the signal was valid. Obviously the waiting period will tend to reduce whipsaws, but it may also give such a late entry that a major portion of a move may missed.
Which Moving Average To Use?
There is no real answer to the question of what combination of moving averages works best. We once saw a matrix, which contained the year-by-year results of every moving average crossover between 1 and 100 going back for 15-years. The conclusion of this study was that moving averages only worked if you knew in advance which combination to use in each commodity each year. Frank Hochheimer did the largest published trading test results we're aware of in the early 1980's at Merrill Lynch.
We have done considerable work in this area ourselves and have tested hundreds of thousands of moving averages. We don't believe there is a magic answer. In practically every case, moving average values which worked well over past data just don't hold up well in present-day trading. This holds true whether they were optimized values or not. In testing, however, and in others where actual trade listings were available, one phenomenon kept reappearing.
As obvious as it sounds, almost any moving average combination is profitable if a market is trending, and almost no combination will produce profits if a trading market is not trending. The answer therefore is not in the search for the perfect moving average. It is in the search for a reliable system that will isolate the markets in which moving averages will be profitable. Then we want to trade those markets in a manner calculated to capture the most profit with least drawdown. Non-trending markets, as we have stated many times before, should be avoided or traded using a counter trend type of indicator.
How To Make a Moving Average "Work"
Moving averages are the simplest and most elegant trend-following study available. Within their limits, they can also be very effective. The limitations of moving average systems, however, can be severe. Most markets spend more time consolidating than they do trending. A non-trending market can be a tough, an even fatal test for the most carefully chosen moving average trading system. Here are some of our thoughts and conclusions on how to help a moving average system survive.
1. Try to confine your trading to only the trending markets. Diversification helps, but don't trade all markets indiscriminately. At any one time, less than 50% of all markets can be defined as trending. Most of the time the actual number is considerably less than that. Find a way to objectively define a market as trending, and only then apply moving averages. We recommend Wilder's DM1/ADX as a reliable study that indicates the extent to which a market is directional or trendless. A simple explanation is when the ADX is rising, the market is trending, and when it is falling, the market is directionless. We also believe the Commodity Channel Index indicator has some applications as a tool for finding and measuring trending markets.
2. Moving averages in general react too slowly to be useful for exits. Use an alternative exit strategy. We think most common mistake made with moving averages is using the same set of moving averages for both entries and exits. If you use slow averages you will do well on the entry side, but will be too slow to exit and give back most of the profits. If you use faster moving averages, you'll have better exits but find yourself getting whipsawed on the trade entries.