Technical Analysis Explained Page 20

  FIGURE 13.23 (a) Directional Changes of Smoothed Momentum MAs. (b) Overbought/Oversold Re-crossovers of Smoothed Momentum MAs

  If the momentum series is found to be unduly volatile, it is always possible to smooth out fluctuations by calculating an even longer-term MA, or by smoothing the MA itself with an additional calculation.

  Another possibility is to construct an oscillator by combining the MAs of three or four ROCs and weighting them according to their time span. This possibility is discussed at length in Chapter 15.

  Chart 13.8 shows the effectiveness of combining two ROC indicators and smoothing them. In this case, the smoothing is a 10-month weighted average of 11- and 14-month rates of change of the S&P Composite monthly closing prices.

  CHART 13.8 S&P Composite 1980–2012 Coppock Indicator

  This was an approach devised by E. S. C. Coppock. Since this indicator has been found useful for market bottoms rather than tops, the momentum curve is significant only when it falls below the zero reference line and then rises. The arrows show that bull market signals between 1982 and 2012 were particularly timely. I have traced this indicator back to 1900 and found that only 3 signals out of 29, in 1913, 1941, and 2002, were premature. Clearly an excellent track record. Even these signals could have been filtered by requiring a positive 12-month MA crossover. Please note that this indicator can be just as easily applied to other markets or relationships.

  The lower panel shows how it is possible to also incorporate trendline violations and price pattern completions using the raw data, i.e., the sum of the two ROCs.

  A further variation on construction of a smoothed momentum index is to take the ROC of an MA of a price index itself. This method reverses the process described earlier, for instead of constructing an ROC and then smoothing the resulting momentum index, the price index itself is first smoothed with an MA and an ROC is taken of that smoothing.

  Summary

  1. Momentum is a generic term embracing many different types of oscillators.

  2. Momentum measures the rate at which prices rise or fall. It gives useful indications of latent strengths or weakness in a price trend. This is because prices usually rise at their fastest pace well ahead of their peak and normally decline at their greatest speed before their ultimate low.

  3. Since markets generally spend more time in a rising than a falling phase, the lead characteristic of momentum indicators is normally greater during rallies than during reactions.

  4. Oscillators reflect market sentiment and have different characters in primary bull and bear markets.

  5. There are two basic methods of interpreting momentum: momentum characteristics and momentum trend reversals.

  6. Momentum signals should always be used in conjunction with a trend-reversal signal by the actual price.

  14 MOMENTUM II: INDIVIDUAL INDICATORS

  It is recommended that you study all of the momentum indicators described in this and the next chapter, subsequently selecting the two or three with which you feel intuitively comfortable with. Following too many indicators will likely lead to confusion. If you wish to find a more indepth explanation of these and many other momentum indicators, please refer to my Definitive Guide to Momentum Indicators (Marketplace Books, 2009) or the momentum module in my online audio/visual course at Pring.com.

  The RSI

  The Formula

  The relative strength indicator (RSI) was developed by Wells Wilder.l It is a momentum indicator, or oscillator, that measures the relative internal strength of a stock or market against itself, instead of comparing one asset with another, or a stock with a market. The formula for the RSI is as follows:

  where RS = the average of x days’ up closes divided by the average of x days’ down closes. The indicator’s design aims to overcome two problems involved in construction of a momentum indicator: (1) erratic movements and (2) the need for a constant trading band for comparison purposes. Erratic movements are caused by sharp alterations in the values, which are dropped off in the calculation. For example, in a 20-day rate-of-change (ROC) indicator, a sharp decline or advance 20 days in the past can cause sudden shifts in the momentum line even if the current price is little changed. The RSI attempts to smooth out such distortions.

  The RSI formula not only provides this smoothing characteristic, but also results in an indicator that fluctuates in a constant range between 0 and l00. The default time span recommended by Wilder is 14 days, which he justified on the basis that it was half of the 28-day lunar cycle. Unfortunately, the lunar cycle includes weekends and, therefore, has more than 14 trading days. Nonetheless, this default time span works quite well in reality, and that’s what really counts.

  The RSI Is Useful for Making Comparisons Between Securities

  The nature of the RSI calculation allows the accurate comparison of different securities on the same chart. In Chart 14.1 there are two series: the Dow Jones Utilities and the Philadelphia Gold and Silver Share Index. The upper panel plots a 45-day ROC, and the lower one a 45-day RSI. With the ROC, it is not possible to easily compare the two because the Utilities are far less volatile. On the other hand, you can see that the divergence in volatility is far less in the lower panel featuring the RSI.

  CHART 14.1 Roc versus RSI characteristics

  Constructing Overbought/Oversold Lines

  Because of this, it is much easier to establish universal standards for the overbought and oversold benchmarks. Using the 14-day default, they are traditionally set at 30 for oversold and 70 for overbought. In an article entitled “How RSI Behaves,”2 Peter W. Aan argued that the average value of an RSI top and bottom occurred close to the 72 and 32 levels, respectively. This research would indicate that the 70 and 30 levels recommended by Wilder should be moved further apart to better reflect the average overbought and oversold value.

  It is important to note that the magnitude of the oscillations of the RSI is inverse to that of most other momentum series. For example, the ROC indicator is subject to wider fluctuations as the time span becomes longer. It works in an opposite way for the RSI. For the RSI, equilibrium is the halfway point, which in this case is the 50 level. It is, therefore, traditional to place the overbought oversold lines equidistant from this point. We should remember that longer time spans in the RSI calculation result in shallower swings and vice versa. Consequently, the 70/30 combination is inappropriate when the time span differs appreciably in either direction from the standard 14-day period. Chart 14.2, for example, features a 9-day RSI for the Hang Seng Index in Hong Kong, where an 80/20 combination gives a much better feel for the overbought/oversold extreme than the 70/30 default value. This is due to the fact that shorter time spans result in wider RSI oscillations. The lower panel features a 65-day RSI where the narrower swings result in a more appropriate 62.5/37.5 combination. In this instance, neither of the default 70/30 values is reached at any time.

  CHART 14.2 Hang Seng, RSI 1999–2010 overbought/oversold Time Span comparisons

  The terms “long” and “short” with regard to time spans refer to the type of data under consideration in a relative sense. For example, a 60-day RSI would represent a long span for daily data, but for monthly numbers, a 60-day, i.e., 2-month, span would be very short. Some consideration should therefore be given to this factor when choosing a specific RSI time span. Chart 14.3 plots two RSIs for identical time periods (60 days or 2 months). However, the overbought/oversold lines are drawn at different levels because one calculation is based on daily and the other on monthly data.

  CHART 14.3 Procter & Gamble 40-day versus 2-month RSI

  Because RSIs based on shorter-term time spans are more volatile, they are more suitable for pointing out overbought and oversold conditions. On the other hand, longer-term spans are more stable in their trajectories and therefore lend themselves better for the purpose of constructing trendlines and price patterns.

  Time Spans

  The RSI can be plotted for any time span. In his book The New Commodity Trading Sy
stems and Methods (Wiley, 1987). Perry Kaufman questions the exclusivity of the 14-day time span (the default) selection. He points out that maximum divergence occurs when the moving average is exactly half the time span of the dominant cycle. In other words, if you make the assumption that the primary trend of the stock market revolves around the 4-year business cycle, a moving average of 24 months will give you the greatest divergence between the high and low points of the cycle. In the case of the 28-day cycle, 14 days is the correct choice, but it is important to understand that there are many other cycles. Working on this assumption, for example, would mean that a 14-hour RSI would be inappropriate if the dominant cycle was something other than 28 hours. The same would be true for weekly and monthly data.

  In practice, a 14-day time span works quite well, but only for shorter periods. I also use 9-, 25-, 30-, and 45-day spans. For weekly data, the calendar quarters operate effectively, so 13-, 26-, 39-, and 52-week spans are adopted. As for monthly charts, the same spans for the ROC are recommended, i.e., 9, 12, 18, and 24 months. For longer-term charts, covering perhaps two years of weekly data, a time span of about 8 weeks offers enough information to identify intermediate-term turning points. A 26-week RSI results in a momentum series that oscillates in a narrower range, but, nevertheless, usually lends itself to trendline construction. Very long-term charts, going back 10 to 20 years, seem to respond well to a 12-month time span. Crossovers of the 30 percent oversold and 70 percent overbought barriers give a very good idea of major long-term buying and selling points. When the RSI pushes through these extremes and then crosses back toward the 50 level, it often warns of a reversal in the primary trend. Remember, these suggested spans tend to work consistently well, but never perfectly well. If you are looking for perfection, technical analysis is probably not your best analytical choice!

  To isolate major buy candidates, it is important to remember that the best opportunities lie where long-term momentum, such as a 12-month RSI, is oversold. In this respect, Chart 14.4 shows that an oversold reversal for the S&P using a 12-month RSI has triggered some very prescient signals in the last 100 years. As with all indicators, this one is not perfect, as you can see from the premature buy signal in the early part of the twentieth century. More care is needed in identifying tops, as equities have a natural tendency to take longer to build than to tear down. Consequently, overbought reversals have a habit of being premature, especially when the secular trend is a positive one.

  CHART 14.4 S&P composite, 1898–2012 12-Month RSI

  Returning to buying opportunities, if you can also identify an intermediate-and a short-term oversold condition, all three trends—primary, intermediate-term, and short-term—are then in a classic position to give a high-probability buy signal.

  RSI Interpretation

  Some of the principal methods used to interpret the RSI are as follows.

  Extreme Readings and Failure Swings Any time an RSI moves above its overbought or below its oversold zone, it indicates the security in question is ripe for a turn. The significance depends upon the time. For example, the 45-hour RSI shown in Chart 14.7 is nowhere near as significant as an RSI constructed with a 12-month time span, as in Chart 14.4. An overbought or oversold reading merely indicates that, in terms of probabilities, a counter-reaction is overdone or overdue. It presents an opportunity to consider liquidation or acquisition, but not an actual buy or sell signal.

  More often than not, the RSI traces out a divergence, as in Figure 14.1. In this case, the second crossover of the extreme level at points A and B usually offers good buy and sell alerts. These divergences are often called failure swings.

  FIGURE 14.1 RSI Failure Swing

  We see a bearish failure swing at the end of 2011 at point A in Chart 14.5 featuring a smoothed version of a 9-day RSI.

  CHART 14.5 Molex, 2009–2011 Smoothed RSI

  Trendline Violations and Pattern Completions The RSI can also be used in conjunction with trendline violations. Generally speaking, the longer the time span for any particular period, i.e., daily, weekly, or monthly, the better the opportunity for trendline construction. Important buy and sell signals are generated when trendlines for both price and the RSI are violated within a relatively short period. Chart 14.6 features a 14-day RSI for Caterpillar. The RSI starts off after a decline by forming a reverse head-and-shoulders pattern, which is more or less simultaneously confirmed by the price. After a good rally materializes, the RSI violates this up trendline. Note how the extended line becomes resistance for the final rally. We can also construct a trendline for the price. Finally, the RSI completes a broadening formation with a flat bottom, and a nasty downside break ensues. The next rally peak also experiences a small RSI top, which is confirmed with a trendline violation by the price.

  CHART 14.6 caterpillar, 2009–2010 RSI and Price Patterns

  An example of the RSI’s ability to form price patterns is shown in Charts 14.6 and 14.7.

  Chart 14.7 shows an hourly chart for Intel. It has a 45-hour RSI in the lower panel, which roughly corresponds to a week of trading. I have drawn the overbought/oversold lines at 62 and 38. See what a nice combination of trendline breaks we get in April 2011. We see another setup in late June, where a trendline break on the RSI is confirmed by one on the price. It’s even possible to observe a right-angled triangle. If you study the equilibrium line at 50, it’s apparent that the strong late June breakout develops slightly above this critical point of balance.

  CHART 14.7 Intel and a 45-Hour RSI Featuring Price Patterns

  Smoothing the RSI It is a perfectly legitimate technique to smooth the RSI. One of my favorite approaches is to smooth a 9-day RSI with an 8-day moving average (MA). Because the fluctuations are not as great as the raw data, the overbought/oversold lines are drawn at 70 and 30, not my usual default of 80/20 for a 9-day span.

  Chart 14.5, featuring Molex, contains a 9-day RSI smoothed with an 8-day MA. The downward-pointing arrows flag all of the overbought reversals. The solid ones show confirmed reversals with varying degrees of success. The dashed arrows indicate overbought reversals where it was not possible to construct a meaningful trendline. Needless to say, they were all failures.

  RSI and Peak-and-Trough Progression The RSI often traces out a series of rising or falling peaks and troughs, which, when reversed, offer important buy or sell alerts. Chart 14.8 shows that the 14-day RSI for Suntrust Banks experienced two peak and trough reversals, each of which was confirmed by a price trend break. These are flagged by the arrows.

  CHART 14.8 Sun Trust Banks, 1997 RSI and Peak-and-Trough Analysis

  Two Variations on the RSI

  Chande Momentum Oscillator

  The Chande Momentum Oscillator (CMO), named after its inventor Dr. Tushar Chande, is a variation on the RSI, yet is uniquely different. It has three characteristics:

  1. The calculations are based on data that have not been smoothed. This means that extreme short-term movements are not hidden, so the indicator reaches overbought/oversold extremes more often, but not enough to result in too many signals.

  2. The scale is confined within the –100 to +100 range. This means that the zero level becomes the equilibrium point. With the RSI, the 50 level is the equilibrium point, and is not always readily identifiable. With zero as the pivotal point, it is easier to see those periods when momentum is positive and those when it is negative. The zero equilibrium, therefore, makes comparisons between different securities that much easier as well.

  3. The formula uses both up and down days in the calculation.

  Interpretation Chart 14.9 compares a 14-day RSI with a 14-day CMO. The first thing to notice is that the CMO reaches an overbought/oversold extreme more times than the RSI—in February and June of 2000 and January of 2001, for instance. Sometimes it is possible to construct timelier and better trendlines for the CMO, though occasionally it works the other way. The March 2000 break, for instance, came off a better trendline for the CMO than the RSI; so, too, did the breaks from trendline AB
and CD. The two trendlines in the summer of 2000 were slightly better for the RSI. Note also that both series experienced positive divergences at the October 2000 low, but that for the CMO was a much stronger signal, since the September bottom was well above that of mid-October. It doesn’t always work in favor of the CMO, but I prefer it because of the more numerous overbought/oversold readings and the plus and minus scaling, which makes it easier to spot positive and negative readings.

  CHART 14.9 FTSE, 2000–2001 comparing the RSI with the CMO

  One approach that I have found helpful is to plot a 20-day CMO and smooth it with a 10-day MA, such as that plotted in Chart 14.10. I then take a smoothing of this indicator—in this case, a 10-day simple moving average. That’s the dashed line that hugs close to the CMO, using the crossovers to generate buy and sell alerts. However, since there are a lot of moving-average crossovers, it’s important to try to filter out those that are not likely to work out. That’s done by taking crossovers that develop at an extreme level more seriously, since they tend to be more accurate. Then make sure that this is confirmed by a trend break in the price. Some examples are shown in the chart.