Technical Analysis Explained Read online

  CHART 14.10 Heng Seng, 1998–2000 Smoothed CMO

  Relative Momentum Index

  The relative momentum index, or RMI, is another variation on the RSI. When calculating the RMI, the standard RSI formula is modified to allow for a momentum factor. The actual formula by Roger Altman was published in the February 1993 Stocks and Commodities magazine article.

  This modification has two effects. First, it smooths the indicator, and second, it accentuates the degree of the fluctuation. The result is a less jagged oscillator that experiences more overbought/oversold readings. The RMI requires two parameters: the time frame and the momentum factor.

  If the RMI has a momentum factor of 1, the indicator is identical to the RSI. It is only when the momentum factor is greater than 1 that the two series diverge. Chart 14.11 shows two variations on the RMI. The middle panel features a 14-day span with an 8-day momentum factor, and the lower one a 45-day span with a 10-day momentum factor. Since it is an RSI-based indicator, longer-term spans involve less volatility. Note that the fluctuations in the 45-day series are much less pronounced than the 14-day RMI.

  CHART 14.11 ATT 1996–2001 Two RMI Variations

  Generally speaking, the longer-term span offers slower and more deliberate movements that lend themselves more easily to trendline construction. Several examples are shown in Chart 14.11, but I particularly like the late 1998 signal since it is confirmed by a simultaneous breakout above the trendline and the 200-day MA. Whenever a price crosses above a trendline and reliable moving average simultaneously, it emphasizes the strength of the signal since they reinforce each other as dynamic resistance areas.

  RSI Conclusion

  Most of the time, the RSI and its two variations, like all oscillators, are not telling us very much. It can be really useful when it triggers divergences, complete price patterns, or violate trendlines. When such characteristics are also confirmed by a trend-reversal signal in the price itself, it is usually a wise policy to pay attention because the RSI has a good record of reliability.

  Trend Deviation (Price Oscillator)

  A trend-deviation indicator is obtained by dividing or subtracting a security’s price by a measure of trend, which is usually a form of MA. It is also possible to base a trend deviation using linear regression techniques. However, we will concentrate on the moving-average technique here. This approach is also called a “price oscillator” in some charting packages. Of the two approaches, subtraction or division, division is preferred since it is more reflective of proportionate moves. For a discussion on this topic, you are referred to Chapters 6 and 8, which compare the logarithmic and arithmetic scales. Chart 14.12 plots the two approaches using a 1/10 price oscillator. The “1” and “10” in the price oscillator legend refer to the fact that a 1-day MA (i.e., the close) is divided by a 10-day MA. Note how the volatility becomes exaggerated as the point movements in the gold price become greater in the last few years. The division calculation returns far more rational swings in the indicator. For short-term charts, where prices do not experience large percentage movements between the low and the high, this is not an important distinction. However, when prices are being compared over many years and there is a substantial net loss or gain over that period, it is wiser to adopt the division approach.

  CHART 14.12 Spot Gold, 1985–2011 comparing Price oscillator comparing Subtraction and Division calculations

  Since the average represents the trend being monitored, the momentum indicator shows how fast the price is advancing or declining in relation to that trend. An oscillator based on a trend-deviation calculation is, in fact, a horizontal representation of the envelope analysis discussed in Chapter 12, but it also shows subtle changes of underlying technical strengths and weaknesses. The top panel of Chart 14.13 shows the price of Brookline Bancorp and its 50-day MA. Two bands at +10 and –10 percent of the 50-day MA have been plotted above and below it. The bottom panel represents the same data but expressed in momentum (price oscillator) format. Therefore, the MA appears as the equilibrium line at zero and the two bands as overbought and oversold levels at +10 and –10 percent. In this way a negative zero crossover is the same as a negative 50-day crossover and so forth.

  CHART 14.13 Brookline Bancorp, 2010–2011 Price oscillator Interpretation

  The interpretation of a trend-deviation indicator is based on the same principles described in Chapter 13. This method can be used to identify divergences and overbought and oversold zones, but it appears to come into its own when used in conjunction with trendline construction and MA crossovers.

  Trendline Construction

  Chart 14.14 shows the crude oil price together with a trend deviation calculated from a close divided by a 45-day MA. This is a fairly jagged indicator and lends itself to overbought/oversold, trendline, and price pattern analysis. We see a good example of a descending triangle break along with a price confirmation in May 2011. Later on, there is barely any upside momentum at point A, and the price and oscillator both subsequently violate trendlines. Most of the time it’s not possible to forecast the character of a price move following a technical event. In this case, though, the very weak positive momentum at A indicated the vulnerability of the situation. Later on, in June 2012, we see an unconfirmed reverse head and shoulders in the oscillator. I say “unconfirmed” because it was not really possible to construct a meaningful down trendline on the price. Finally, a price pattern completion for the oscillator and a trendline violation for the price leave the chart with a confirmed short-term downtrend.

  CHART 14.14 Light crude, 2011–2012 Price oscillator Interpretation

  Trend Deviation and MAs

  An alternative approach with trend deviation indicators is to smooth out unwanted volatility with the aid of two MAs, as shown in Chart 14.15. The actual trend-deviation series is calculated by taking a 26-week MA of the closing price divided by a 52-week MA. The second series is simply a 10-week MA of the first. Buy and sell alerts are then triggered as the smoothed trend-deviation indicator crosses above or below its 10-week MA. Then look for a confirmation from the price itself. Two examples are shown in Chart 14.15, one for a top and the other for a bottom. This is very much a guerilla approach because the buy alert indicated by the dashed arrow was signaled almost at the top of the rally. This example demonstrates the importance of picking and choosing between signals, only selecting those that develop close to a turning point. If this filtering approach is not taken, then there is considerable risk that action will be taken close to the end of the trend.

  CHART 14.15 S&P Airlines, 1995–2001, and a Smoothed Trend-Deviation Indicator

  A useful method that greatly reduces such whipsaw activity but still offers timely signals is to lag the 52-week MA by 10 weeks when the trend-deviation calculation is being made. This means that each weekly close is divided by the 52-week MA as it appeared 10 weeks before. This new calculation has been plotted in the center panel of Chart 14.16.

  CHART 14.16 S&P Airlines, 1995–2001, and Two Smoothed Trend-Deviation Indicators

  In this example, the whipsaw in late 2000 was filtered out since the trend-deviation indicator fails to cross decisively below its MA. I am not suggesting this is the only legitimate combination for weekly charts, but it is one that appears to operate quite well. There is always a trade-off when you try to make signals less sensitive, and in this case, we find that there is occasionally a small delay compared to the nonlagged 52-week MA. The most obvious one on this chart developed at the beginning of 1997, where the lagged series in the center panel crossed its MA at a slightly higher price. In most instances though, this is a small price to pay if a costly whipsaw can be avoided.

  MACD

  The moving-average convergence divergence (MACD) trading method is form of trend-deviation indicator using two exponential moving averages, the shorter being subtracted from the longer. The two MAs are usually calculated on an exponential basis in which more recent periods are more heavily weighted than in the case of a simple MA.
It is normal for the MACD to then be smoothed by a third exponential moving average (EMA), which is plotted separately on the chart. This average is known as the “signal line,” the crossovers of which generate buy and sell signals. It obtains its name from the fact that the two EMAs are continually converging and then diverging from each other. The MACD has gained great popularity over the years, but in effect, it is really just another variation on a trend-deviation indicator that employs two EMAs as its method of deviation. A visual of its construction is therefore very similar to Chart 14.13.

  MACDs can be used in an infinite number of time periods. Gerald Appel of Signalert,3 who has done a considerable amount of research on the subject, recommends that buy signals on a daily chart be constructed from a combination of 8, 17, and 9 exponential MAs, but he feels that sell signals are more reliable when triggered on the basis of a 12, 25, and 9 combination. On the other hand, the popular MetaStock program plots the default values as 12 and 26 with the signal line at 9.

  Chart 14.17 shows Microsoft in 2011 and 2012. The dotted arrows indicate whipsaw signals, and the long arrows indicate those signals that were confirmed with a trendline break of some kind.

  CHART 14.17 Microsoft, 2010–2011 MACD Interpretation

  General Electric is shown with an MACD indicator in Chart 14.18. Another technique is to construct overbought/oversold lines, trendlines, and price patterns and to look for divergences. In Chart 14.18, for instance, both series complete head-and-shoulders patterns at the end of the year 2000. The MACD also experiences a negative divergence. Note how the divergence, flagged by the right shoulder, is barely able to rally above zero. The result is an above-average decline. Note also that the indicator remained below the equilibrium point and touched its oversold level several times during the balance of the period covered by the chart. This type of action reflects bear market activity.

  CHART 14.18 General Electric and an MACD Indicator

  The MACD often works well with monthly data. In this respect, we look at the CRB Spot Raw Industrial Commodity Index in Chart 14.19, where the solid arrows indicate good primary-trend momentum buy signals. The two dashed ones indicate smaller rallies, which fall into the failure category.

  CHART 14.19 CRB Spot Raw Industrials, 1977–2012 Long-Term MACD Buy Signals

  The MACD is often plotted in a histogram format and the signal line against it, as in Chart 14.20 featuring Homestake Mining. The chart shows a classic head-and-shoulders pattern. Note that the MACD histogram gradually became weaker as the pattern progressed. This was only a short-term sell signal, but the price eventually fell below the signal level.

  CHART 14.20 Homestake Mining and an MACD in Histogram Format

  Stochastics

  The stochastic indicator has also gained a great deal of popularity among futures traders, with the result that the standard formula uses very short-term time spans. The theory behind the indicator, which was invented by George Lane,4 is that prices tend to close near the upper end of a trading range during an uptrend. As the trend matures, the tendency for prices to close away from the high of the session becomes pronounced. In a downward-moving market, the reverse conditions hold true.

  The stochastic indicator, therefore, attempts to measure the points in a rising trend at which the closing prices tend to cluster around the lows for the period in question, and vice versa, since these are the conditions that signal trend reversals. It is plotted as two lines: the %K line and the %D line. The %D line is the one that provides the major signals and is, therefore, more important.

  The formula for calculation of %K is:

  %K = 100[(C – L5close)/(H5 - L5)]

  where C is the most recent close, L5 is the lowest low for the last five trading periods, and H5 is the highest high for the same five trading periods. Remember that the calculation of stochastic indicators differs from that of most other momentum indicators in that it requires high, low, and closing data for the calculation.

  The stochastic formula is similar to the RSI in that the plots can never exceed 0 or 100, but in this case, it measures the closing price in relation to the total price range for a selected number of periods. A very high reading in excess of 80 would put the closing price for the period near the top of the range, while a low reading under 20 would put it near the bottom of the range.

  The second line, %D, is a smoothed version of the %K line. The normal value is three periods. The %D formula is as follows:

  %D = 100 × (H3/L3)

  where H3 is the three-period sum of (C – L5) and L3 is the three-period sum of (H5 – L5).

  The momentum indicator that results from these calculations is two lines that fluctuate between 0 and 100. The %K line is usually plotted as a solid line, while the slower %D line is usually plotted as a dashed line.

  The popularity of the stochastic indicator can no doubt be explained by the smooth manner in which it moves from an overbought to an oversold condition, lulling a trader into feeling that price trends are much more orderly than would appear from an observation of an RSI or an ROC indicator.

  Longer-term time frames, used on monthly and weekly charts, appear to work much better than the shorter-term stochastics used on daily futures charts. Colby and Meyers, in The Encyclopedia of Technical Market Indicators (McGraw-Hill, 2002),5 noted that the stochastic indicator tested very poorly relative to MA crossovers and other momentum indicators.

  Overbought and oversold bands for the stochastic are usually plotted in the 75 to 85 percent area on the upside and in the 15 to 25 percent area on the downside, depending on the time span in question. An overbought indication is given when the %D line crosses the extreme band, but an actual sell alert is not indicated until the %K line crosses below it. When the two lines cross, they behave very similarly to a dual MA system. If you wait for the penetration, you can avoid getting trapped into shorting a strongly bullish move or buying an extremely negative one.

  The behavior of the indicator will depend very much on the selected time frames. Since the %K and %D can be thought of as moving averages, it follows that the longer the time span, the smoother the resultant indicator. The upper panel of Chart 14.21 shows a very volatile 5/5 combination, whereas the lower window contains a 30/100 combination. Note that the %K is much less volatile than that in the upper window. The same is true for the 100 parameter used in the %D. However, it is so flat that it is essentially useless from the point of view of reflecting the cyclic rhythms. In reality, one would never use such a combination.

  CHART 14.21 S&P composite, 2010–2011 comparing Two Stochastic Time Spans

  Slowed Stochastic

  It is also possible (and desirable) to extend the calculation in order to invoke a slowed version of the stochastic. In this instance, the %K line is replaced with the %D line, and another MA is calculated for the %D. Many technicians argue that this modified stochastic version gives more accurate signals. It certainly results in a more deliberate action. A comparison between a regular and a slowed stochastic is featured in Chart 14.22. The 5/5 in the upper window legend refers to a 5 period for the %K and D, respectively. The 5 in the middle of the lower window legend refers to the slowing factor.

  CHART 14.22 S&P Composite, 2010–2011 Comparing a Stochastic with a Slowed Stochastic

  Chart 14.22 compares a regular stochastic to a slowed one.

  General Interpretation

  Crossovers Normally, the faster %K line changes direction sooner than the %D line. This means that the crossover will occur before the %D line has reversed direction, as in Figure 14.2.

  FIGURE 14.2 Stochastic crossovers

  Divergences Figure 14.3 shows examples of where the %K fails to confirm a new high or low in the price, thereby setting up a divergence, which when confirmed, signals a change in trend.

  FIGURE 14.3 Stochastic Positive and Negative Divergences

  Divergence Failure An important indication of a possible change in trend arises when the %K line crosses the %D line, moves back to test its extreme l
evel, and fails to cross the %D line, as in Figure 14.4. Once again, it is always important to see some price confirmation, and that is given with the break above the down trendline in the right side of the diagram.

  FIGURE 14.4 Stochastic Divergence Failure

  Chart 14.23 shows some examples of divergence failures: two confirmed bearish ones and one unconfirmed bullish one

  CHART 14.23 Microsoft and Divergence Failures

  Reverse Divergence Occasionally during an uptrend, the %D line will make a lower low, which is associated with a higher low in the price, as shown in Figure 14.5.

  FIGURE 14.5 Stochastic Negative Reverse Divergence

  This is a bearish omen, and conventional wisdom suggests looking for a selling opportunity on the next rally. This condition is sometimes referred to as a bear setup. A positive reverse divergence is featured in Figure 14.6.

  FIGURE 14.6 Stochastic Positive Reverse Divergence

  Extremes Occasionally the %K value reaches the extreme of 100 or 0. This indicates that a very powerful move is under way, since the price is consistently closing near its high or low. If a successful test of this extreme occurs following a pullback, it is usually an excellent entry point.