Trading Strategies Based On Digital Filters - page 65

 

GOERTZEL spectrum wihtout detrending ?

richcap:
Hi all,

while I was lurking I played a bit with Goertzel.

The pictures below should clarify that, with stationary (or quasi-stationary) AWGN-noised cyclic signal, both MESA and Goertzel give robust and very similar results.

Enjoy ;-)

Did you get GOERTZEL spectrum without detrending ? From MATLAB it was not possible, i had to use detrend function to recover spectrum.

Swords seems to be similar, i know. Extra funcionality of Meyers is just to make reverse transform on a few frequencies with the highest amplitude and make some phase advancing. Similar is done Goertzel V2

Krzysztof

 
fajst_k:
You dont have to dream....look to this, but its done in different way.

https://www.mql5.com/en/forum/178285

post 123 and down.

Actually market signal looks quite different than our sin+cos+noise+trend signal, its a sum of groups of subcycles which most likely by luck are in sync at certain moment. You can see this looking at eigenvectors using NOXA

show eigenvector tool. Maybe I will post the picture.

Krzysztof

This thread doesn't have 123+ post. Total of 77 as of this post. Perhaps you are using SSA to forecast future post?

Just kidding......

But seriously, to what were you referring?

 
forex_for_life:
This thread doesn't have 123+ post. Total of 77 as of this post. Perhaps you are using SSA to forecast future post?

Just kidding......

But seriously, to what were you referring?

sorry it was post 73. 123 was the number of my posts.

 
richcap:
You are right.

What I really mean is that if we (mainly Krzysztof) want to tell MESA from Goertzl (from SSA), the test signals we have dealed so far are useless.

We should analyze non-stationary, non AWGN added, 'test' timeseries.

Maybe a sequence of known multi tone "tunes" with some sort of multiplicative noise.

We should try to add noise with our experience as traders. I mean, if in a timeseries synthetized from intermittent sine/cosine waves of various frequencies, we see a human readable support/resistance, we would expect the timeseries to move in S/R-aware way.

A S/R cannot change a strong trend but it introduces noise in it.

We should try to model this behaviour somehow...

OK stop dreaming, I return back to my lurking...

You dont have to dream....look to this, but its done in different way.

https://www.mql5.com/en/forum/178285

post 73 and down.

Actually market signal looks quite different than our sin+cos+noise+trend signal, its a sum of groups of subcycles which most likely by luck are in sync at certain moment. You can see this looking at eigenvectors using NOXA

show eigenvector tool. Maybe I will post the picture.

Krzysztof

 

measuring cycle with filter banks

Here are two screenshots how the latest Ehlers method of measuring cycle with

filter bank works. It seems works fine. Code attached in pdf

Krzysztof

Files:
fb1.jpg  115 kb
fb2.jpg  122 kb
 

Dominant cycle finding

Hi Richcap, Simba

In this paper there is a code and description of Ehlers way finding dominant cycle using Center of Gravity algorithm. Perhaps better than user changable number of cycles to be used in inverse transform as it is now in Goertzel V2. So Simba you can improve your sword if you want as it will become more adaptive.

Krzysztof

 

cycle analysis

Here is cycle analysis of three pairs. I used for this some not available for MT4 indicators but cycles are cycles. out of sample test starts with green bars and it was set to 48h so last Friday and Thursday.

First EURUSD. From H4 you can see that there is no cycle anymore (period > 30 bars) but rather trend, from H1 you can see that period length is about 20 bars and swing position goes up so it is a long sign. Strategy is sligthly profitable out of sample on H1&H4 since 48h

Files:
h4eu.jpg  95 kb
h1eu.jpg  110 kb
 

Gbpjpy

Here GBPJPY went to declining trend line and it depends if the break will be succesfull or not it will take a direction. On H1 there is already a sell signal but swing position is already very low. So the best to wait for result of the test of this trend line

Files:
h4gy.jpg  101 kb
h1gy.jpg  113 kb
 

Gbpusd

It is also trending not cycling on H4, on H1 swing position and CSSA cycles looking for a rise so most likely continuation of the uptrend.

Comments are welcome. All those strategies were quite much profitable out of

sample for last 48h with PF 3.08 on H1 and hit 100% on H4.

Krzysztof

Files:
h4gu.jpg  94 kb
h1gu.jpg  106 kb
 
fajst_k:
Here is cycle analysis of three pairs. I used for this some not available for MT4 indicators but cycles are cycles. out of sample test starts with green bars and it was set to 48h so last Friday and Thursday. First EURUSD. From H4 you can see that there is no cycle anymore (period > 30 bars) but rather trend, from H1 you can see that period length is about 20 bars and swing position goes up so it is a long sign. Strategy is sligthly profitable out of sample on H1&H4 since 48h

K,

Thanks for posting your analysis,I will post my comments on them once I exactly understand the meaning..BTW,don`t see this as confrontational,just the opposite,I have been mistaken more times than rigth,so,the issue is not to compete,but translate a method into tradable "signals/directions",and then compare relative value.

Is your view on EURUSD H4+H1 ..Long?No cycle,trend UP?

Same for GBPUSD?

GBPJPY undecided?bouncing up and down in a triangle?

Regards

Simba