16 replies to this topic

[da_cheif]

naw....not u......me......geeziz

[jjc]

Why do I waste my time here?!?

Ah, coma kisa. Don't give up on us.

"cyclical analysis uses the latest information"

I'm not smart here about the "cyclical analysis" that you refer; However the fact that your relying on
cycles as a predictor means you are relying on historic information. Without twice the wavelength
of data your results are meaningless (our friends Nyquest and Shannon). In fact if higher frequencies exist in your data sequence your going to end up seeing a bunch of false harmonics.

[arbman]

Any technical analysis is essentially separating out the momentum and cyclicality, so no guarantees in any of them, so it is better to be informed. I have a model based on the dominant periodicities that might contain all of the harmonics in them at once to minimize the noise and maximize the correlation rather than separating into the individual harmonics. I use only about 3 periodicities backward so as to minimize the variation since I do not correct for variation separately. However, I do use the advection of the cyclical patterns by using a 2D Taylor polynomial of 3rd degree. It explains the blow off and crash like extremes much better, it must be people acting as they see the patterns emerging, I guess. In a sense, it is never cyclical analysis or expectation of an exact cyclicality, it is just recent patterns that repeat and influence each other in my system. The closest to my model is the periodicity transform (look up on Google), I posted here before too. Once I have the dominant periodicities, scanning for the historical similarities to evaluate the statistical odds is still better than the EW woodoo. The other measures such as the breath momentum, support and resistance get evaluated separately in my model, I iterate in different ways, it is all numerical molding anyway... - kisa

[jjc]

Very interesting. You lost me on your advection use. Your 2D poly is carrying Price and Volume info? Price and Volatility? How many matches does your historical scans typically yield? Have you tried it on smaller time scales were the historic data is plentiful? Thanks for the quick lesson. I've been meaning to read that paper you posted... just have too many irons in the fire.

[arbman]

This is only using the prices, yet as it is "a tool" of analysis, it can run on anything with historical price, money flow etc. The model can make forecasts of its own, but because it is a legitimate model of "flow", you can also find the highly correlating statistical cases having the same flow characteristics (periodicity and shape) and compare. I have a few more models where I sort and analyze by market caps, sectors etc. The Taylor series with multiple variables can accomodate infinite number of variables as long as you know how the derivatives of each variable relate to each other, for example the general fluid flow (Navier-Stokes) equation can be derived from it. I simply modeled a flow derived from the decomposition of the price patterns among the highest correlating ones by using the finite differences among the patterns. There is no real cyclical expectancy in my model, it just happens. It is an autocorrelation model similar to the Fractals without scaling the time... This is enough detail though...

[SilentOne]

Why do I waste my time here?!?

kisacik,

For what its worth, I agree with your statement.

"The cycles are more predictive than EW"

I like EW, but I have come to the conclusion that understanding the market cycles is more important. EW takes a back seat. JMHO of course.

cheers,

john

Edited by SilentOne, 05 May 2007 - 05:37 AM.

[da_cheif]

Why do I waste my time here?!?

kisacik,

For what its worth, I agree with your statement.

"The cycles are more predictive than EW"

I like EW, but I have come to the conclusion that understanding the market cycles is more important. EW takes a back seat. JMHO of course.

cheers,

john

if you only understand english......i wouldnt be surprised if russian didnt take a back seat either....