|
PureBytes Links
Trading Reference Links
|
"Michael J. Doiron" <75534.2410@xxxxxxxxxxxxxx> wrote:
> I guess I'm unclear of the definition of velocity. Momentum is
> showing me the change in price, but not necessarily "how fast" it
> is - which is what I assumed velocity to mean. But that could be
> my trouble - an assumption.
I suspect a lot of people (myself included) are confused here, partly
because of what seems to be incorrect terminology. As Trade Jack
noted, velocity and momentum are basically interchangeable. Correct?
In physics, velocity is the change in position between two points in
time. In trading, e.g. in TradeStation, *momentum* is the change in
price between two points in time. Sure sounds like velocity to me.
But since "momentum" is the standard terminology, we can use that.
Momentum (velocity) is the derivative of price / position.
Acceleration is the derivative of Momentum (velocity).
So if Momentum is Price - Price[M], then I would think Acceleration
would be Momentum - Momentum[N]. Correct?
So Momentum measures "how fast" the price is moving. This seems to
be useful, but you'd also like to know how fast the MOMENTUM is
changing -- which is the acceleration.
Why is this of interest? Because the Momentum crosses zero at market
turning points. When the market is moving up, Momentum is positive;
when it's moving down, Momentum is negative. If the Momentum is
changing at a constant rate, then you can predict WHERE the market
will turn, which would obviously be useful information. Constantly
changing momentum means a constant Acceleration. Unfortunately the
market very seldom has a constant Acceleration, so I'm not entirely
sure how to work with the changing Acceleration. (Do we need to take
another derivative of Acceration to see how *it's* fluctuating??
Seems like you could keep this up to the Nth degree until you delay
the signal beyond any usefulness...)
I said above that momentum crosses zero at market turns. That's
true, but only for the *instantaneous* momentum. As Mark Jurik said
recently, there really isn't such an animal -- you have to compare
two prices to compute momentum, so by definition you must have lag.
In order to reduce noise, you need to increase lag, and then the
whole value of momentum (& thereby acceleration) as a turn indicator
decreases. You can improve the results by smoothing the price with a
low-lag filter like Jurik's AMA or Tillson's T3, but I'm still not
quite sure how to really make use of this.
If I chart this, Momentum and Acceleration actually seem to track
each other fairly closely. At least, if I plot Momentum(C,5) and
Momentum(C,5)-Momentum(C,5)[5], they track. If I use a shorter time
span on the Acceleration calculation it seems a bit more useful, but
of course it's more noisy.
I guess a lot of my difficulty stems from not understanding exactly
how to apply this so it actually produces useful information. Right
now the plots I see are much more noise than usable content.
Mark B or others, could you spend a moment (or several :-) to clarify
for us how you actually use this?
Thanks,
Gary
|