TABLE 34.2 Treasury Bonds Price Oscillator Optimization Results, 1981–1998
On the face of it, the number of losing signals of 234 to 131 winners looks pretty grim. However, when you look at the more detailed report of Table 34.3, the average win was 2.2 times greater than the average loss, which shows this system did a reasonable job of cutting losses short.
TABLE 34.3 Treasury Bonds Using the 28/2/-2 Combination, 1981–1998
The top panel of Chart 34.4 shows the equity line. The starting amount of $1 was increased to $2.55. Even though the system trailed the buy-hold approach, there were no major drawdowns in terms of peak-to-trough equity. The one in 1994 of 10 percent was the worst. Not bad, considering the 150 percent gain was achieved at a 9.4 percent annualized rate.
CHART 34.4 Treasury Bonds, 1981–1998, Featuring the 28/2/–2 Price Oscillator System
Other tests on many closed-end mutual funds covering the 1980s and 1990s have shown that the 28/5/–5 combination of a 28-day MA and a close divided by a 28-day MA using +5 and –5 as the overbought/oversold triggering points worked consistently well.
The Triple Indicator System
One important principle that should be followed when designing a system incorporating several triggering mechanisms is to make sure it incorporates different indicators based on different time frames. The contrasted time frames are important because prices at any one time are determined by the interaction of many different time cycles. We cannot make provisions for all of them, of course. If we can ensure there is a good time difference that separates the indicators used in the construction, we will at least have made an attempt to monitor more than one cycle.
A system I devised in the late 1970s combines an MA crossover with a signal from two rate-of-change (ROC) indicators. These are a 10-week simple MA, a 6-week ROC, and a 13-week ROC. Thus, we have two different types of indicators: a trend-following MA and two oscillator types. The system also consists of three different time frames. The buy and sell rules are very simple. Buy when the price is above the 10-week MA and both ROCs are above zero. Sell when all three go negative, that is, when the ROCs cross below zero and the price crosses its MA. Signals cannot be generated unless all three agree. This is because we want to make sure the various cycles reflected in the three different time frames are all in gear. Originally, when I introduced this system, it was applied to the pound/dollar relationship because it was one that trended very consistently.
Let’s take a close look at Chart 34.5 to see how it works by starting off with a simple 10-week MA crossover between mid-1974 and 1976. Buy signals are once again indicated by the upward-pointing arrows and short positions by the downward-pointing ones. There were 13 signals for a total profit of $0.19 on an initial $1 investment, from both the long and the short side. This compares to the buy-hold approach, with a loss of almost $0.70. Taken on its own, this was a fairly commendable performance, but let’s remember that for a significant portion of the time, that is, most of 1975 and 1976, the British pound was in a sustained downtrend. It is true there were a number of whipsaws in late 1975 and early 1976. These are shown in the two ellipses, but they were of minor consequence as it turned out.
CHART 34.5 British Pound System and a 10-Week MA
The next step is to introduce a 13-week ROC. Buy and sell signals are triggered when the 13-week ROC crosses above and below its zero reference line. This approach, shown in Chart 34.6, nets a gain of $0.23 with six signals. This was better than the results with the MA crossover, especially since fewer signals dramatically reduced the potential for whipsaws. Even so, there were a couple of nasty whipsaws in 1976.
CHART 34.6 British Pound System and a 13-Week ROC
The next step is to introduce a second ROC indicator to filter out some of the whipsaws. A 6-week ROC was chosen mainly because it spans approximately half the time span of the 13-week series. The result was an improved $0.24, but the number of signals increased to 12. The 6-week ROC is shown in the middle panel of Chart 34.7.
Putting the Indicators Together
I put all three indicators together in Chart 34.7 so you can see how their integration improves things. The actual result was a slight increase in profit over the previous 6-week ROC test. However, the important thing was that the signals were reduced to only three. A closer look at Chart 34.7 shows that the first sell comes in October 1974, as the 6-week ROC follows the others into negative territory. Then, in December, the 13-week ROC crosses above zero, and this is closely followed by an MA crossover. Finally, the 6-week series moves above zero for a buy signal. All three then move into negative territory in April 1975. The MA and 6-week ROC go bearish simultaneously, and this is then followed by the 13-week series. The system stays bearish all the way through late 1976. It almost goes bullish when the price crosses its average and the 6-week ROC goes positive in February 1976. However, the 13-week series, which had been bearish, now goes bullish, but by this time, the currency had slipped below its MA and the 6-week series fell below zero. As a result, all three indicators were never in agreement. The same is true in the July–August period of 1976 when the two ROC indicators take turns in being bullish and bearish. This was a type of a negative complex divergence described in Chapter 13. The combination of all three indicators works extremely well in this environment.
CHART 34.7 British Pound System and Three Indicators
This is about as good as it gets.
Appraising the System
I originally introduced this approach in my book International Investing Made Easy (McGraw-Hill, 1981) with some hesitancy because there was obviously no guarantee it would continue to operate profitably. It was subsequently reintroduced in the third edition of this book in 1992 with the same proviso. What I said was, “It is important to understand this approach will not necessarily offer such large rewards in the future. The example of the British pound must be treated as the exception rather than the rule, but it is introduced to give you an incentive to experiment along these lines.”
The system continued to work extremely well, as you can see by looking back at the equity line in the upper panel in Chart 34.8. However, I am glad I used the cautionary statement, because once we move past 1993, the system fell apart. Just look at the declining equity line between 1993 and 2002 in Chart 34.8. Indeed, though it made money for the next 10 years, the peak 1993 equity has never been surpassed. The initial drop was due to the many whipsaws arising from the trading range that followed the drop from $2.00 in 1993. This goes to show that even if a system works well for 20 years, as this one had, market conditions can and do change, so you must be prepared for such instances. Obviously, we do not know until sometime after the fact that the market environment was different. Is there anything we can do to avoid such situations? One possibility is to run a very long-term MA or trendline through the equity line.
CHART 34.8 British Pound System Results, 1983–1998
In Chart 34.9, I have plotted a 300-week simple MA against the equity line of the three-indicator pound system when applied to the Hang Seng Index. I used 300 weeks because I felt it necessary for the system to undergo a fairly long period before it can be considered out of touch. The idea is when the equity line crosses below the MA something is seriously wrong with the system, and it should be at least temporarily abandoned. At this point, it would make sense to reappraise it and see if it could be improved, and I do not mean by introducing special rules to block out a bad period. You could also wait until the equity line crosses back above the MA again, but in the case of the “busted” pound system, as far as the pound was concerned, even that approach was problematic. Unfortunately, such eventualities are unavoidable, which is why it’s a good idea to diversify among securities and systems in order to reduce risk.
CHART 34.9 British Pound System Applied to the Hang Seng Index, 1981–2012
Introducing an Intermarket System
The Relationship
So far, we have just considered particular securities or mark
ets in isolation, using statistically derived data from that security alone. An alternative approach is to adopt a tried and tested intermarket relationship as a cross-reference.
Better results are often obtained in this way. An intermarket relationship develops when one market consistently influences another. The first step is to rationalize why such a relationship exists in the first place. Perhaps the most basic one is between equities and short-term interest rates. This was described in Chapter 31, where it was established that changes in the trend of short-term interest rates lead equity prices.
What we do not know is the lead time or the magnitude of the ensuing stock rally. The answer is to classify the trend of money-market prices, which is what the inverted short rate actually is, with an MA crossover. When a rising trend of money-market prices has been established, it is then time to look at the trend of equities to see when they respond. The rationale is that a rising trend of money-market prices sets the scene for an equity bull market. However, this is not confirmed until the S&P Composite crosses above its MA. Just think of this as something akin to an unconscious swimmer receiving mouth-to-mouth resuscitation. You know the treatment is good for the patient, just as falling rates are good for equity prices. However, we do not know how much treatment is required and whether the patient will recover until he or she is able to breathe by him-or herself. In our analogy, the stock market is shown to respond to the interest-rate treatment when it crosses its MA.
Here is how it works. Look at Chart 34.10. In October 1981, the inverted commercial paper yield crosses above its 12-month MA (shown in the ellipse), indicating the environment is now bullish for equities. However, the equity market does not respond by bottoming out until August 1982. When the S&P rallies above its 12-month MA (A), it indicates that the market is responding to the positive interest rate environment. In this case, the crossover comes in August 1982. At that time, both trends are bullish and so is the system. It remains positive until either series moves back below its average, which, in this case, developed in June 1983 (B). It then goes bullish again in January 1985 (C).
CHART 34.10 S&P Composite, 1980–1988, versus 3-month Commercial Paper Yield
Finally, the inverted yield falls below its average in early 1987 (D). The market continues to rally, but the system is no longer bullish. In most instances, it would be better to generate the sell signals after the S&P crosses below its average. In this instance, though, the 1987 crash was over before the average was penetrated. Since the risk increases as the money-market series crosses below its average, it is probably best to act on the signal in two parts. This would involve taking off half the position as the money-market series goes negative and then liquidating the rest when the S&P crosses its average.
Figure 34.5 compares the risk and reward for the system between 1900 and 2009 to that for the S&P Composite. The vertical axis measures the monthly reward on an annualized basis, and the risk is measured on the horizontal axis. In this sense, risk is measured as volatility. The best place for any system to be is in the top-left corner, often referred to as the northwest quadrant. This is where the reward is high and the risk, or volatility, is low. In the case of this system, you can see that the risk was slightly lower than that for the S&P, but the reward was substantially higher. At Pring Turner Capital, we call this the 120 percent rule because it has had such an excellent and consistent track record for over 100 years.
FIGURE 34.5 The 120% Rule Risk versus Return
The system says nothing about periods when the market is above its average and rates are not, since those are obviously bullish periods as well. However, once rates move above their 12-month MA, there is a real danger that the next correction could be the first downleg in a bear market. It is true that sooner or later the S&P Composite will cross below its MA, thereby stopping us out, but why run the risk when good returns and little volatility can be had under more favorable conditions? During those periods when both were in a negative mode, the annualized loss was 9.58 percent.
If you are a short-term trader, you probably feel this approach is worse than useless. However, it can be put to very good use if you realize that when the system is bullish, short sell signals are more likely to result in losses. They are not just going against the main trend, but are occurring in one of the most positive equity environments you can get. By the same token, this knowledge can be used to take positions yourself on the long side when a short-term buy signal is triggered. I am not going to say that sharp corrections will never happen when this system is positive, because there have been periods such as 1971 when a fairly large retracement move did materialize. It is merely that when the system is positive, the odds favor strong short-term rallies and whipsaw reactions.
Using Margin
All the systems described here were tested on a cash basis with no margin. You might think that it makes sense to go out and apply a system using lots of margin. That way the gains would multiply. In actual fact, this is not necessarily the case. Chart 34.11 shows a simple 10-day MA crossover system using no margin, and Chart 34.12 shows it using a 10 percent margin requirement. Since the initial trades were losers, the account was wiped out in just over a year. Remember: Leverage works both ways.
CHART 34.11 American Century Gold Fund, 1989–1998
CHART 34.12 American Century Gold Fund, 1989–1998
Summary
1. There are two ways in which systems can be used: act on each signal without question, or use the signals as a filter so the current system becomes one more indicator in the weight-of-the-evidence approach.
2. The principal advantage of a mechanical system is that it removes subjectivity and encourages the adoption of discipline.
3. No system will ever work all of the time. It is important to understand the pitfalls of automated systems so that they can be programmed out.
4. Systems should be designed to take account of the fact that there are two different types of market environment: trading range and trending.
5. Because no system works perfectly, it should be exhaustively tested before being applied to the marketplace.
6. The use of any system should involve diversification to spread the risk for any period where it does not operate successfully for a specific security.
7. Incorporating a tried and tested intermarket relationship into the system acts as a cross reference and usually enhances results.
8. The use of margin exaggerates the results, both on the upside and the downside. The actual performance will depend on the chronological sequence of the good and bad signals.
35 CHECKPOINTS FOR IDENTIFYING PRIMARY STOCK MARKET PEAKS AND TROUGHS
Primary bull market tops and bottoms are elusive affairs, largely because the points at which we expect them to develop are the ones that appear to be the most unlikely at the time. When most people are lucky enough to identify a bull market peak, they assume that prices will immediately decline. This is not normally the case because of the numerous and confusing cross-currents necessary for the development of true distribution. This topping-out process usually requires a trading range environment, which reflects the tremendous battle between bulls and bears. First one comes out on top and then the other. By the time the distribution has been completed, both sides are exhausted. Even though the bears eventually win, most lose their original conviction because of the numerous false rallies that they did not anticipate. These advances develop under an environment of extreme optimism, which also make them all the more convincing to those caught short or who have already sold out.
We know that the news is favorable at market peaks, but when a typical top is staring us in the face, the widespread contagion of optimism deludes us into expecting conditions to get even better. The opposite is true of bottoms: The news is bad, but we expect it to get worse before prices hit their low. Alternatively, we expect the other shoe to drop, just like traumatized earthquake victims anticipate a killer aftershock. If it seems inconceivable, given the prevailing conditions
and likely outlook, that prices will rise, they probably will.
As an initial step, and this would apply to any security, it’s a good idea to make an attempt to figure out the current position of the cycle using monthly data calculated from your favorite long-term smoothed momentum curve, e.g., the Know Sure Thing (KST), stochastic (20/14/10), moving-average convergence divergence (MACD), etc.
Though by no means infallible, these series have a strong tendency to reverse direction around the time of a primary peak or trough. If you spot a reversal of or some observe stalling action, it’s then a good idea to revert to evidence provided by other indicators.
The Mechanics of a Peak
A classic market peak should involve a battle between early- and late-cycle leaders. As the top begins, liquidity-driven issues peak out and begin their bear market (see Figure 35.1).
Technical Analysis Explained Page 51