Random Flow Theory and FOREX - page 8

New comment
 

Yes, I've already seen it. :-)

 
lna01: They are of course periodic, but with period of first 2^n more signallen+patternlen, exactly up to this length zeros are added - it follows from the source. That is in fact non-periodic :)
Yes, I did not look into the code, I'm not very fond of this business. But the problem is that the ACF (in this case the AC matrix) is a function depending on two arguments in the general case of a non-stationary process. And here, no matter how you look at it, it's one dimensional.

I'll have to see what a correlogram is, thanks, Neutron. Maybe that's what it's all about...
 
Mathemat:
But the problem is that the ACF (in this case the AC matrix) is a function depending on two arguments in the general case of a non-stationary process. And here, as it were, it's one dimensional.

I'll have to see what a correlogram is, thanks, Neutron. Maybe that's what it's all about...

The historical panorama will also require a 3D image.

As far as I understood Neutron's code considers the same as fastcorellation, but for the limit case of a single point pattern. By the way, I can't tell right away which of the procedures will be faster for this limit.

P.S. No, wrong, the code is not for a point but for a pattern in NBars, that is just the same as fastcorellation, then fastcorellation should be faster.

 

As long as you are doing statistics like this, you might as well get acquainted with standard time series models, especially since this is a piece from an econometrics course: http://education.iet.ru/files/text/econometrics/lect/econometrics_lecture_02.pdf. The text seems to be easy to understand.

 
Mathemat:

Well, as long as you're doing statistics like that, you might as well get familiar with standard time series models...

That's our way! Otherwise... pilots, planes... - you know!

Just kidding.

The pilot in an aeroplane has one big problem - the aeroplane cannot change its position in space by leaps and bounds, i.e. the change of target coordinate, is a continuous and smooth function. The functions belonging to this class have positive autocorrelation function, so even if we have target signal laden with parasitic noises with known distribution laws, it is not very difficult to determinate useful signal. Even elementary smoothing will do, since smoothing works on this class of functions, and all the apparatus of classical functional analysis will work. It will not be difficult to predict with a given accuracy the future values of such a function...

Unfortunately, for BP of the type of financial instrument quotes, having, as a rule, the sign-variable correlogram and negative autocorrelation coefficient on all TFs - everything sucks (in terms of prediction). No smoothing, BFT etc. will work here by definition :-(( Regression analysis gives such an opportunity. However, everything is complicated here.

 

Neutron

Let me disagree with some of your conclusions. Yes the aircraft has mass and that helps. But have you tried to predict the movement of an aircraft in a turbulent atmosphere when it can at any moment make any manoeuvre from a stationary flight, combat turn,..... to the famous cobra, And all measurements occur in space and in polar coordinates + you yourself are still moving and the enemy puts you interference such that the GEP with a hairpin on quotations is a lekkha fun?

And everything is on a plane (if we add Volum is three-dimensional), the Cartesian coordinate system + we are not moving + we have time to analyse as much as necessary. In combat modes, sorry :(

I am attaching a file with an example and explanations of how to analyze flow and build ACF. If the curve view is saved on the entire history for all currencies (only parameters will be changed), it will be a light at the end of the tunnel leading to the grail, IHMO.

I'm just trying to do everything right (naturally from my point of view), because if you do x... But it may work ;-).

Files:
akf_3.zip  103 kb
 
Prival:

I just try to do everything right (naturally, from my point of view), because if you do x... But it might work ;-).


I agree with you!

Here is the link to the full version of Mathemat' s article: http://window.edu.ru/window/catalog?p_mode=1&p_rid=39071&p_rubr=2.2.76.4. 8

The material is of high quality.

 

I will post the ACF figure here to immediately assess what I have got there. Note the function is smooth, everything is based on real quotes.

Edit.

Thanks for the links, I will do my best to read them all, wish I had more time to sleep :-(. I have been working 3 jobs. Maybe at least one of them should be an analyst at the brokerage house. If you are such a fool as I need, just call, under certain circumstances, I will quit all three jobs.

 
lna01:

2 Prival:

I couldn't resist and made some cosmetic changes to the indicator code

I have some kind of problems, it is not shown completely, and it only appears on М1 and М15. The quotes archive is updated. The refresh chart menu does not work :-(

 
Prival:

I have some errors, it is not shown completely, and it only appears on M1 and M15. I attach the image, the quotes archive is updated. The menu item update chart does not help :-(

Is there any record in the log about division by zero? This indicator has no protection against such a situation.

P.S. I've looked at the same area myself - it does. I attach a variant with protection.
Files:
pvr42xcs_1.mq4  4 kb
123456789101112131415...85
New comment
Reason: