THE HAndbook of TECHnICAl AnAlysIsthat is, A/(A + D + U). Another breadth indicator that uses total issues in its
denominator is the High Low Logic Index, created by Norman Fosback.- Accumulation: Sometimes it is useful to represent raw data as a daily running
total. We can do this by simply adding the daily operator results to the previous
day’s running total. For example, the cumulative form of the daily net advances
would simply be: (A − D) + Previous Day’s Running Total Net Advances. Popu-
lar breadth indicators that use the cumulative form include the Advance Decline
Line, New High New Low Line, and the McClellan Summation Index. - Smoothing: The resulting values obtained by various raw data operations may
frequently be too visually erratic or volatile. This erratic line may be smoothed
by applying a moving average to the resulting operator values. Normally a
simple moving average is sufficient to smooth the erratic operator readings,
that is, (∑ (A − D) over last N days)/N. It is not uncommon to employ weighted,
double, or triple smoothing to the resulting operator values. - Crossover Triggering: In order to reduce the lag in the smoothed operator val-
ues caused by the application of single moving averages, the crossover of two
moving averages is sometimes employed. This results in less erratic readings. - Detrending: Many breadth indicators are designed to isolate the difference
between the values of two moving averages by subtracting the values of
one moving average from the other, for each successive period. Such opera-
tions are termed detrending. One popular breadth indicator that employs the
detrending of two exponential moving averages is the McClellan Oscillator.
(The McClellan Oscillator is in fact mathematically identical to the MACD
[moving average convergence‐divergence, except for the two lookback periods
and the data used.) - Normalization: Many oscillators are unbounded, that is, the indicator values
have no upper or lower limit. Many of these indicators may be transformed
into bounded indicators via the process of normalizing. Although unbounded
oscillators may still indicate periods of overbought or oversold by compar-
ing highs and lows with historically significant levels, bounded oscillators
sometimes provide a simpler and more objective representation of such over-
extensions, where readings near or above 80 percent represent overbought
conditions and those near or below 20 percent represent oversold conditions.
Normalizing involves subtracting the lowest value over the last N periods
from the current or closing value and dividing this by the difference between
the highest and lowest values across the last N periods. (This is in fact the
basic mathematical construct of George Lane’s Stochastic Oscillator.) For ex-
ample, we could design a normalized Advance Decline indicator such as in
Figure 22.1.
figure 22.1 Normalizing Breadth Data.
Current (A−D) − [Lowest (A−D) over the last N period]
[Highest (A−D) − Lowest (A−D)] over the last N period