|
PureBytes Links
Trading Reference Links
|
no sore point then since i posted my reply to you in private. now
that you dragged it out into public, i'm compelled to reply.
i asked some simple questions about momentum and acceleration. period.
both mark jurik and bob fulks replied on target, and answered my
questions. you, however, did not. the only "sore" point (as you call
it)i have is wading through a bunch of verbage regarding your t-bond
trading system and not finding a clear explanation to the original
momentum questions that i had asked.
that's basically the gist of my complaint with you.
TJ
here are my comments to a mark jurik reply re: momentum
thanks mark!
now that i see where you're going with the argument, i tend to agree.
also, i think that the various authors i cited may in fact mean that
momentum "leads" price in the sense that it "diverges" from price, ie,
momentum changes direction before price changes direction. some cases
it does, and other cases it doesn't.
whatever
TJ
thanks again
---Mark Jurik <mgj@xxxxxxxxxxxx> wrote:
Yes, you can create very low lag momentum measurements. I originally
spoke of a 7-bar *simple* momentum indicator as having 3 days lag.
However, more advanced techniques can push the lag even lower, without
raising the noise level. And that's the key tradeoff, lag versus
noise. The best indicators have low levels of both.
The pure derivative is a coincident indicator because it represents
the instantaneous momentum (tangent) to a continuous curve. It has no
lag. However, in the real world of noisy, discrete time signals, there
is no instantaneous momentum. You have to approximate it by
evaluating at least two samples, and one of those must be historical.
The farther they are apart, the less noisy they become, but the
greater the inherent lag you must contend with. So, for practical
reasons revolving around noise reduction, momentum is a lagging
indicator. As to why some refer to it as a leading indicator, I guess
it's because taking the first derivative (momentum) of a sine wave
creates
a cosine, which appears to lead the sine perfectly by 1/4 wavelength.
However, that is a special case where it happens to lead well. If you
apply this test to the more general case of a random walk, momentum
doesn't lead squat. Since the market sometimes trends and sometimes
not, then sometimes it may appear to lead and sometimes not. I don't
find that view very useful.
For me its better to acknowledge the limitations of indicators and any
corresponding home-brew maps into the future that I may have generated
from them. This way. I'm conscious of the map applied, and can
question its validity.
The phrase "zero-lag" refers to the fact that it smoothes momentum
without adding any lag whatsoever in the smoothing process. This
permits you to lower the length (and lower the lag) while maintaining
acceptable noise rejection. Again, there's always that trade-off.
|