|
PureBytes Links
Trading Reference Links
|
i mostly agree with what you say. i think you can play with momentum
to produce an indicator that has consistently very low lag, ie, what
you call a "coincident" indicator.
but why do many authors, including Kaufman, Blau, even Chuck LeBeau
(in his book), claim that momentum (and indirectly acceleration) are
"leading" indicators? bob fulks also argued that under some
conditions, momentum can be a "leading" indicator, as well as a
"lagging" indicator. and isn't your "vel" thingy a "coincident
indicator."
what am i missing here? is it just a question of semantics and all
indicators are lagging, or some lag less than others?
TJ
---Mark Jurik <mgj@xxxxxxxxxxxx> wrote:
For me, a lagging indicator measures some past feature of a time
series. A coincident indicator measures some current feature, and a
leading indicator
measures some (anticipated) future feature.
Momentum is not a coincident indicator. If I had a coincident
indicator that gave me the instantaneous average momentum of any
market, I'd become wealthy very fast. Unfortunately, a 7-bar simple
momentum, for example, tells you the average momentum of the market 4
bars ago. Not the momentum today. Therefore, it is a lagging
indicator. This goes for acceleration measurements as well.
A forecast requires mapping one or more lagging and coincident
features into the future. The *assumption* that yesterday's momentum
will persist 1 or more additional bars is a basis for mapping momentum
into a leading indicator. For example, linear regression is a lagging
momentum indicator, but its extrapolation can produce a leading
indicator. Whether or not
extrapolation creates anything useful in a trading system needs to be
tested.
Pattern recognition (eg. head-and-shoulders) is a lagging indicator,
because by the time you are confident it is present, most of it
already occurred. The belief that such patterns complete themselves
with high probability is the basis for mapping a forecast, which
becomes the leading indicator.
|