CHART 27.2 The DJIA and the Long-Term A/D Line, 1931–1983
A/D Lines Using Daily Data
Because daily A/D lines have a tendency toward a downward bias, some care should be used in comparing recent highs with those achieved 2 to 3 years ago. Daily A/D lines come into their own when they fail to confirm new highs in the market average that have occurred within an 18-month period. An example is shown in Chart 27.3a, where the A/D line peaks in April 1987 but the S&P Composite does not top out until late August. The S&P did not fall right away, but eventually followed the leadership of the A/D line. Quite often, a number of divergences will be set up. Initially, these might be well publicized, but since the widely expected decline fails to materialize, many technicians give up, stating that the divergence “won’t work this time.” Invariably it does work, though much later than most would anticipate. This was very much the case at the market peak in January 1973, which was followed by a 2-year divergence.
CHART 27.3a The S&P Composite versus the Daily NYSE A/D Line, 1986–1988
Because bottoms in the daily line usually coincide with or lag behind bottoms in the average, they are not very useful at this point for the purpose of identifying a trend reversal.
A more practical approach is to construct a trendline for both the A/D line and the market average. Violation of both lines usually signals that an important rally is under way. Some examples are shown in Chart 27.3b.
CHART 27.3b The S&P Composite versus the Daily NYSE A/D Line, 1991–1995
Two resistance lines are violated at the end of 1992. Later on, the two dashed up trendlines are violated for a joint sell signal. Note that in this case the lines are penetrated at approximately the same time that both series cross below their respective 200-day MAs. This joint evidence adds to our weight-of-the-evidence approach and increases the odds of a valid breakout. Finally, both series violate down trendlines at the beginning of 1995.
When considering potential divergences or nonconfirmations, it is always important to give these relationships some room. For example, at point A it may have appeared at the time that the A/D line was going to experience a major negative divergence since it had yet to beat its early 1994 high. It would have been easy to jump to a bearish conclusion. However, this would not have been supported by the facts since the A/D line was well above its 200-day MA at this point. Moreover, there was no sign of a trend break in the S&P Composite that would confirm the negative divergence even had it existed. As it turned out, both series went on to make significant new highs, thereby pointing out the importance of giving the prevailing trend and this relationship the benefit of the doubt.
Charts 27.4 and 27.5 also compare the S&P Composite to the daily A/D line. The first shows a negative divergence that opened up at the 2007 bull market peak. I have highlighted the 2009 low, though this was not an actual divergence, as both simultaneously registered new lows, though that for the breadth indicator was clearly less intense. This slight discrepancy did represent a small warning that things might turn to the upside, but really required the kind of confirmation given at point B, where two down trend-lines were penetrated on the upside. The action at A was a positive divergence as the A/D line experienced rising bottoms but not the price. This was again confirmed with two nice trendline violations, and a solid rally followed. Chart 27.5 shows a later period. First we see a confirmed negative discrepancy at the July 2011 intermediate top. This time, it’s different because the S&P registered its high ahead of the line. It really does not matter which way this disagreement falls, though usually it comes from A/D line weakness. The key is that when both series confirm with a trendline break or reliable MA crossover, a discrepancy is a discrepancy and prices fall. Points A and B offer two examples of a rare situation whereby the A/D line bottoms ahead of the S&P. In the case of A, we see some actionable confirmation. However, in B’s case, the drop was so steep that it was not possible to construct timely trendlines that could have served as actionable technical events.
CHART 27.4 The S&P Composite versus the Daily NYSE A/D Line, 2007–2009
CHART 27.5 The S&P Composite versus the Daily NYSE A/D Line, 2010–2012
Breadth Oscillators (Internal Strength)
For historical comparative purposes, the rate-of-change (ROC) method of determining momentum is useful in measuring price indexes because it reflects moves of similar proportion in an identical way. This method, however, is not suitable for gauging the vitality of the indicators constructed from cumulative data that monitor internal market structure, such as those that measure volume or breadth. This is because the construction of such indexes is often started from a purely arbitrary number. Under certain circumstances, this might require an ROC to be calculated between a negative and a positive number, which would obviously give a completely false impression of the prevailing trend of momentum. The following sections provide a brief summary of some oscillators constructed from breadth data using a more suitable method of calculation.
Ten-Week A/D Oscillator
Chart 27.6 shows the DJIA and a 10-week oscillator calculated from a 10-week MA of the square root of the A/U – D/U formula discussed earlier. A comparison of the A/D line to the DJIA illustrates the principle of divergence, as evidenced by declining peaks of momentum and rising peaks in the Dow at the 2007 peak. This was later confirmed by a break below the dashed up trendline. A positive divergence was also confirmed in the spring of 2009. Note that it was possible to establish different ranges for the oscillator depending on the primary trend environment. These are shown by the dashed parallel lines. Also, note the extremely high reading in the oscillator as it came off the 2009 low. This was a mega overbought and represented an early bird warning of a reversal to a bull market. You can also see an extreme but lower reading in the spring of 2003. That was also a mega overbought, but this time signaling the start of the 2003–2007 bull market.
CHART 27.6 The S&P Composite versus the 10-Week A/D Ratio, 2002–2013
Ten- and Thirty-Day A/D Oscillators
These indicators are calculated by taking a 10- or 30-day MA of the A/D or the A – D ratio. An alternative calculation can be made by dividing the total of advancing issues by the total of declining issues over a specific time span. Their interpretation is exactly the same as with other momentum indicators, bearing in mind their relatively short time span. An example of a 10-day breadth momentum series is shown in Chart 27.7.
CHART 27.7 The NYSE Daily A/D Line, 1999–2001, and Two Breadth Oscillators
Note that this time we are comparing the oscillators to the A/D line itself rather than the S&P or DJIA. Both series experience a set of positive divergences between 1999 and March 2000. Then we see some negative divergences as the A/D line peaks out later that year. Note how the 10-day series is barely able to rally above the equilibrium point, indicating extreme weakness at the time of the actual rally high in September. Finally, both the 30-day oscillator and the line itself both violate up trendlines for a classic weight-of-the-evidence sell signal. One final negative divergence develops in January 2001.
The McClellan Oscillator
The McClellan Oscillator is a short-term breadth momentum indicator that measures the difference between a 19- and a 39-day exponential moving average (EMA) of advancing minus declining issues. In this respect, it is based on the same principle as the moving-average convergence divergence (MACD) indicator discussed in Chapter 14. The generally accepted rules are that buy signals are triggered when the McClellan Oscillator falls to the oversold area of –70 to –100 and sell signals are triggered when it rises to the +70 to +100 area. Since the calculation is based on a subtraction method and the number of NYSE issues has grown over the years, these bands are probably too narrow to be of practical use. My own experience suggests that its interpretation should be based on the same principles as those described in Chapter 13 using divergences, trendline analysis, and so forth. An example is shown in Chart 27.8 using breadth data from the NASDAQ exchange. Note the p
ositive divergence that developed at the 2009 low. The dashed vertical arrows flag three important peaks—all were associated with a very low reading in the oscillator. At the 2000 peak, the indicator was actually in slightly negative territory, clearly an extreme reading for an extreme chart point.
CHART 27.8 NASDAQ 100 ETF, 2007–2010, and the McClellan Oscillator
Finally, the oscillator has been described here using two specific time frames for the EMAs for the calculation since these are the generally accepted default values. However, there is nothing to stop the innovative technician from experimenting with different combinations of EMAs.
The McClellan Summation Index is a derivation of the McClellan Oscillator and is calculated as a cumulative total of the daily readings of the oscillator itself. The result is plotted as a slow-moving curve that changes direction whenever the raw oscillator (described earlier) crosses above or below its zero line. The slope of the summation curve is determined by the difference between the actual reading and the zero line. In other words, an overbought reading will cause the summation index to rise sharply, and vice versa. Many technicians use these changes in direction as buy and sell signals, but this can result in a lot of whipsaws. My own preference is to use an MA crossover. This is often less timely, but it filters out a significant number of false signals. A suggested time frame for this exercise is a 35-day simple moving average. An example is featured in Chart 27.9. Even here we see numerous whipsaw signals indicating that this approach is far from perfect.
CHART 27.9 S&P Composite, 1998–2001, and the McClellan Summation Index
High-Low Data
The popular press and many online data providers publish daily and weekly figures for stocks reaching new highs and lows. These statistics relate to the number of issues making new highs or lows over a 52-week period. There are various methods of measuring the high-low figures, but since the raw data are very jagged, displaying them in an MA format is usually better. Some technicians prefer to plot an MA of the two series individually, others an MA of the net difference between highs and lows.
Major Technical Principle A rising market, over a period of time, should be accompanied by a healthy, but not necessarily rising, number of net new highs.
When the major averages trace out a series of higher peaks following a long advance but the net number of new highs forms a series of declining peaks, this is a warning of potential trouble. This type of relationship indicates that the technical picture is gradually weakening because successive peaks in the market average are accompanied by fewer and fewer stocks making breakouts (new highs) from price patterns. The net number of new highs also takes into consideration stocks making new lows. In a bear market, a new low in the S&P Composite or other market average that is not accompanied by a declining number of net new highs is a positive sign.
In this case, a declining number of stocks reaching new lows implies fewer downside breakouts, i.e., a shrinkage in the number of stocks resisting the downtrend in the major averages. In Chart 27.10, for instance, the S&P falls to approximately the same level in December 1994 as it did earlier in the year, yet the number of new lows was far less. This indicated an improving technical position that was eventually confirmed when the index rallied above the solid trendline.
CHART 27.10 The S&P Composite, 1993–1996, and 52-Week New NYSE Lows
The bottom panel in Chart 27.11 shows a 10-day MA of the daily high/low differential. Note the negative divergence between this series and the average between 1989 and 1990, and also the fact that it was possible to construct a couple of (dashed) trendlines for the ratio and the S&P that were violated in early 1991. The implied trend of expanding net new highs was signaling that once the index itself responded with a breakout, prices were likely to move higher.
The Cumulative Net New High series in the second panel is constructed by cumulating the daily difference between the new highs and lows in a similar fashion to the daily A/D line. For example, if there are 100 new highs and 20 new lows, the difference, i.e., 80, would be added into the total and vice versa.1 I have found that using 100-day MA crossovers offers reasonably good signals of when the environment is positive or negative for the overall market. Signals of this nature generated between 1988 and 1993 are indicted by the solid perpendicular arrows in Chart 27.11.
CHART 27.11 S&P Composite, 1988–1993, and Two Net New High Indicators
Chart 27.12 shows more recent price action where the arrows point up the 100-day MA crossovers of the cumulative line.
CHART 27.12 S&P Composite, 2006–2012, and Two New High Indicators
Note that in early 2009 and late 2011 this developed more or less simultaneously with the 200-day MA crossovers of the S&P Composite. An alternative method of calculating high-low data is shown at the bottom of Chart 27.13, where an 8-day MA of net weekly new highs has been plotted against the S&P Composite. The light highlights indicate when this indicator falls below zero. Its main claim to fame is protection against a primary bear market, which it did very well in the 2007–2009 period. However, the drawback is that it often falls into negative territory at the end of a protracted short-term decline.
CHART 27.13 S&P Composite, 2005–2012, and an 8 Day New High Indicator
In this discussion we have limited ourselves to 52-week periods for the new high-low calculations. However, there is no reason why such calculations cannot be made for any time period or any basket of securities. For example, Chart 27.14 shows a cumulative line derived from the net new highs of a basket of commodities calculated over a 165-day time span.
CHART 27.14 Dow Jones UBS Commodity ETN, 2005–2012, and a Commodity New High Indicator
The shaded areas represent when this series crosses above its (dashed) 100-day MA. It’s by no means a perfect indicator, but does offer a view as to whether commodities are in a primary bull or bear market. Finally, Chart 27.15 shows a price oscillator calculated from a cumulative line of a basket of gold stocks. The legend explains the calculation. First, the line is calculated using a 65-day time span—in other words, the number of gold stocks registering net new highs over a 65-day period. The 15 and 65 indicate that the oscillator is calculated by dividing a 15-day MA of the cumulative line by a 65-day MA. As you can see, overbought/oversold reversals offer timely buy and sell alerts. We have used gold shares in this example, but there is no reason why this analysis cannot be extended to other sectors or markets.
CHART 27.15 Gold Miners ETF, 2007–2012, and a New High Oscillator
Diffusion Indicators
A diffusion indicator is a form of oscillator constructed from a basket of items that measures the number or percentage of that universe that are in a positive trend. An example might be the percentage of the 30 stocks comprising the DJIA that are above their 30-day MAs. When all members are in a bullish mode, the picture is as positive as it can get. The implication is that the aggregate measure, the DJIA in our example, is vulnerable and, therefore, likely to peak out. The reverse set of conditions, in which none of the series is in a positive trend, produces the opposite effect; i.e., the aggregate index may be reaching its low point and could, there-fore, be a “buy.” This simple interpretation of diffusion indexes is a good starting point, but in practice, a diffusion measure is a form of momentum indicator, and is subject to the same benefits, drawbacks, and principles of interpretation outlined in Chapter 13.
What Is a Positive Trend?
In technical analysis, a market or stock that forms a series of rising peaks and troughs, or is above a trendline, may be classified as being in a positive trend. However, the only way trends can be monitored through this interpretation is on the basis of individual judgment, which would make the construction of a diffusion index covering many series over many years a very laborious process. For this reason, and because of the need for greater objectivity, a statistical measure that can easily be calculated on a computer is normally used.
The most common measurements calculate the percentage of a series that ar
e above a specific MA or that have a rising MA. Another popular alternative is to take the percentage of a universe of series that have a positive ROC, i.e., a reading above 0 or 100. The choice of the time span for the MA or ROC is very important. The shorter the span, the more volatile the resulting oscillator.
In practice, it seems that the MAs and ROCs commonly used in other areas of technical analysis offer superior results. These are 30-day and 50-day for short-term trends; 13-, 30-, and 40-week for intermediate-term trends; and 9-, 12-, 18-, and 24-month for longer-term trends. The same exercise could also be accomplished with intraday data. One characteristic of using any raw series is that the resulting data usually needs to be smoothed. For example, the diffusion series shown in Chart 27.16 is calculated from the percentage of a basket of commodities and commodity indexes that are above a 24-month MA. This data, in turn, have been smoothed, and thus the solid line actually represents a 9-month MA of the percentage of groups above their respective 24-month MAs. The dashed line is a 9-month MA of the solid (24/9) series. The arrows show that timely primary-trend buy and sell signals are triggered as the diffusion indicator reverses direction and crosses its 9-month (dashed) MA.
CHART 27.16 CRB Spot Raw Industrials, 1969–2012, and a Diffusion Indicator
How Many Items Should Be Included?
Technical Analysis Explained Page 40
