• Sign up >>

Requests & Ideas - page 39

MrM
271
MrM  

contrary to popular opinion there are causal wavelets:

i.e. the Causal Mexican Hat Wavelet

Read Chapter 5 p44of "mathematical techniques in financial market trading"

The book is a bit dated, but then again you are talking wavelets

also: the a trous: http://www.unige.ch/fapse/mad/renaud/papers/PredictMultiscale.pdf

hughesfleming
273
hughesfleming  

One thing that surprised me yesterday when I did some tests with rapidminer which has DFT and DWT was that different broker feeds can give you quite different results. The differences also show up using Richcap's instant spectrum (Maximum Entropy) which is an excellent tool by the way.

My primary feed is Dukascopy but I have been testing an MT4 currenex feed and the differences were quite large. If we do get to try some new cycle tools in the future it might be handy to use a standard set of test data that we could share. Just and idea.

Alex

Rikers
12
Rikers  
hughesfleming:
One thing that surprised me yesterday when I did some tests with rapidminer which has DFT and DWT was that different broker feeds can give you quite different results. The differences also show up using Richcap's instant spectrum (Maximum Entropy) which is an excellent tool by the way. Alex

Generally speaking more robust system won't be much affected by the deviations. What timeframe where you using? Daily data are different from broker to broker because of the GMT offset but deviations from broker to broker on 5x time increment data shouldn't be so different. If they are then it is unusual.

hughesfleming
273
hughesfleming  

It was daily data for EURUSD as that is the only timeframe I use for cycles. I do look for cycles on 4H but only when they confirm what I am seeing on D1. I am usually looking for cycles between 140-180 days and one in particular that I have been following for the last 12 months. When I have some time later, I will look again and try and figure out what the reason was. Nevertheless, broker feed seems to also affect the end result more than I had expected.

Alex

SIMBA
2078
SIMBA  

WAVELETS-Links&Books

mladen:
MrM Even though I like the FilterPeriods presentation better, there is no guarantee that even within the period bars values are going to stay stationary (here is one quote of HP filter :
So, according to that and since lambda conversion to periods is an approximation of the bars, there is no guarantee that it will stay "as is" even out of filter period bars back. Will check this more thoroughly (frankly, I did not spend time enough with HP filter to know all of it) and if I find some different info, will get back to this theme

_____________________________

Now I have one plea too

I found this site : The Wavelet Digest :: Index (even though their forum is spammed by Russian spammers misusing the fact that the forum admin obviously does not know Russian, it has some very useful info - like the Gallery->Tutorials page with links that leads to a wealth of very serious information on the theme)

My plea is if anyone knows something similar, please share the link here.

regards

Mladen

Wavelets and Signal Processing

The Wavelet Packet Transform

Daubechies Wavelets in Java and C++

BOOKS: Unzip with 7zip

An Introduction to Wavelets and Other Filtering Methods in Finance and Economics .djvu password:ebooksclub.org

Link: request download ticket | ifile.it

Scaling, Fractals and Wavelets .pdf password:ebooksclub.org

Link: request download ticket | ifile.it

S

MrM
271
MrM  

eptf Goertzel

Mladen, can you code an eptf version of Crodzilla's famous Goertzel browser?

And maybe also in the work that you are doing right now on the fft end-point you could use the concept of sorting cycles by decreasing amplitude. It's not the VFFT concept but I guess paying 70 bucks doesn't get you an entry ticket to the real party... if you're not inside, you're outside

Simba, MrTools, Mladen,

The Mister Bean reference in the post that you couldn't delete just made me realize that it must be one of you 3 that was using the moniker "FxTimer" who made those posts in that FF thread.

I realize that you will not release the code for the indicator, but por favor, give me a pointer here: besides the references on VFFT & multi-scaling you made there is there any other work that you used to create that indicator? It's obviously the way you combine the waves that is the trick. Do you have an mql4 code example illustrating the multi-scale approach so I can figure it out by myself?

SIMBA
2078
SIMBA  
Mladen Rakic
163387
Mladen Rakic  

MrM

This is just the first step

It roughly removes the problem it had (the problem with not calculating correctly when it was left to work as is - on a picture there is an example of the original (upper) and the changed (lower) and the error that was made starting from the vertical line (an error of not calculating correctly) as well as the values before the vertical line (those were values left after the initial calculation - so they are not value for current values)
It will take some more time since it has to be cleaned up thoroughly before end-pointing it (a lot of reasons, but I really would like to revise the whole thing first, there are some points that might cause serious errors in calculations) so, please treat this just as a first (little) step

PS: it calculates now on every tick (unlike the previous)

regards

Mladen

MrM:
Mladen, can you code an eptf version of Crodzilla's famous Goertzel browser?

And maybe also in the work that you are doing right now on the fft end-point you could use the concept of sorting cycles by decreasing amplitude. It's not the VFFT concept but I guess paying 70 bucks doesn't get you an entry ticket to the real party... if you're not inside, you're outside

Simba, MrTools, Mladen,

The Mister Bean reference in the post that you couldn't delete just made me realize that it must be one of you 3 that was using the moniker "FxTimer" who made those posts in that FF thread.

I realize that you will not release the code for the indicator, but por favor, give me a pointer here: besides the references on VFFT & multi-scaling you made there is there any other work that you used to create that indicator? It's obviously the way you combine the waves that is the trick. Do you have an mql4 code example illustrating the multi-scale approach so I can figure it out by myself?
Files:
MrM
271
MrM  
SIMBA:
And,believe me,70 bucks give you a lot more than we got codeveloped working for months,In these 2 exclusive sections I have found both extraordinary concepts and indicators that may add an edge...

It's true, Mladen, Igorad & MrTools are excellent coders. I'm not looking for a holy grail because I know it's a waste of time, but I think there are techniques out there that are not in the public domain that are better than the ones here. I am a professional trader and I come here to learn more about applied quantitative analysis. "Indicators" do make a difference. Off course it's what you do with all the information that makes the p&l, but I'm not here to learn to trade.. I'm here for good indicators I guess I'm asking for an "ultra advanced elite". Such a section would contain indicators that you, mladen & mrtools use in their MT4 for actual live market trading. Like the one you developed with your by now famous team of 7. My point is that I'm not looking for an edge of 51%. I have that. I'm looking for a world-class idea.

I have resorted on several occasions in the past to rentacoder, ... and they are way more expensive,so,I decided to just ask Mladen or Tools ,and give them their time to code it,besides being exceptional programmers ,they are traders too,so they can understand perfectly well our needs,and,sometimes,even anticipate them....IMHO.

RAC is expensive and this is cheaper. And yes, I did use them. I also said I'd be willing to share the function, just like I'm sharing the Matlab file that UPS delivered at my door 10 minutes ago.

Files:
MrM
271
MrM  
mladen:
It will take some more time since it has to be cleaned up thoroughly before end-pointing it (a lot of reasons, but I really would like to revise the whole thing first, there are some points that might cause serious errors in calculations) so, please treat this just as a first (little) step

Great I look forward to this one.