Chartmill Value Indicator, Part 2 by Dirk Vandycke
Under- And Overvaluation
In the first part of this series, you were introduced to an oscillator
not prone to stickiness in overbought/oversold regions and
that didnít have the lag commonly associated with oscillators.
In part 2, hereís how to apply it and profit from it.
When using the Chartmill value indicator (CVI), which is
a short-term oscillator, I try to capture overbought and
oversold setups. Based on a normalization procedure
well-known to engineers and statisticians, the indicatorís
construction makes for far different properties than those
of most popular oscillators. To begin with, the CVI doesnít have the stickiness that keeps other, range-bound oscillators in
overbought and oversold zones while strong trends develop.
Second, the lag typical for moving averageĖbased oscillators
is far less of an issue with the CVI. Finally, because of its
adaptable dynamic nature, there are no parameters involved
in its equation, so the CVI is totally objective by definition.
But it is still necessary to give it an objective interpretation.
Letís find out if this indicator is of any real and statistical
In my previous article I introduced the CVI and how it is
built. Because of its normalizing nature, it can be used in the same way for any time series. Any value between ‑4 and +4
indicates a strong short-term consensus regarding current
pricing as fair value. Any value between 4 and 8 indicates
slight overvaluation, while anything higher than 8 indicates
plain overpricing in the short term.
Even though the CVI isnít a range-bound oscillator, you
will rarely see values higher than 8 with this indicator. Given
that values for the CVI are next to never much higher than
9, this would indicate major overvaluation with respect to
short-term consensus. In order to cope with the stickiness
effect that plagues other oscillators, the CVI has a tendency
to time-correct toward zero and considers new price extremes
more normal as time goes by.
Likewise, minor undervaluation, in regard to short-term
consensus, is defined as between ‑4 and ‑8 and plain undervaluation
below ‑8. Any value lower than ‑9 would be considered
to be a major undervaluation.
These levels are not written in stone, of course, but they
come from a statistical normalization procedure. As a consequence,
all four beacons pinpoint the CVI equivalent of the
statistical standard deviations.
If you want to find an edge the CVI offers, you should look
for violations of the 8 and ‑8 levels. So your focus here will be
on over- and undervaluation signals. Going still farther from
zero would result in far too few signals to build a meaningful