Window Oscillators and Overlay Indicators
or oversold for extended periods, and as such provide little useful infor-
mation other than that a strong trend is in effect.
B. Unbounded oscillators: The values or readings of unbounded window os-
cillators have no upper or lower limit. Note that although unbounded win-
dow oscillators have no theoretical upper or lower limit, some unbounded
window oscillators do have a lower limit of zero, such as volume, ADX,
Bollinger Bandwidth, and average true range (ATR). These window oscil-
lators are therefore referred to as being semi‐bounded. The main advan-
tage of employing an unbounded oscillator is that it provides historical
levels of overbought and oversold. These historical levels of overbought
and oversold are specific to a particular stock or market at a selected
timeframe, and care should be taken when used to compare the degree
of overextension between different stocks or markets. An unbounded
window oscillator may sometimes be converted into a bounded window
oscillator via normalization, confining the range of window oscillator ac-
tion between 0 and 100 percent. Normalization will always convert a
point‐based oscillator (e.g., MACD, momentum, CCI, ATR, ADX, etc.)
into a percentage scaled oscillator, ranging between 0 and 100 percent. It
should be noted that some practitioners refer to unbounded oscillators as
window indicators, rather than oscillators. This is mainly due to the fact
that an unbounded oscillator need not actually revert back to its central or
equilibrium value. Values can rise or fall indefinitely with no indication of
whether they will revert to equilibrium levels. As such, many practitioners
prefer to use the term indicator as many unbounded oscillators tend to
not oscillate around their equilibrium levels but rather trend in one or the
other direction. Unbounded oscillators that tend to trend include those
that employ a net running or cumulative total of all past values, such as
the Net Advance Decline Line, OBV, and so on.
See Figure 8.1 for comparisons between bounded and unbounded window
oscillators. Notice that overlays may be applied to window oscillators as well.
See Figure 8.2 for a flow chart summary of indicator classification.Chart scale sensitivity and Invariance
Chart scaling only affects indicators that are geometrically based. As such, it does
not affect window oscillators, as the values of all window oscillators are numeri-
cally determined, that is, mathematically calculated. It also does not affect over-
lays that are numerically and algorithmically determined. For example, a simple
moving average represents the average of N period closing prices, irrespective of
whether the chart was scaled arithmetically (linear based) or logarithmically (ratio
based). The average price of N period closing prices is scale invariant, that is, it
does not change with scaling. Another example of scale invariance involves algo-
rithmically determined overlays. Assume that we wanted to buy at the breach of
a five‐day high. Once we have identified such a pattern, we would initiate a long
position once price breaches the five‐day high. Again, we see that such sequence
bar patterns are unaffected by scaling. The high prices will still have the same