Building a trading system using digital low-pass filters - page 10

New comment
 
olyakish:
NorthernWind:
olyakish:


Yes, there is a basis for this explanation. The main thing is that this "crowd effect" has a place in the graph.

The most interesting thing is that it can be used in trading ...

And in fact it's the most interesting thing in 11 !!! pages.
 

Here we are talking so much about stationarity and other clever things. But there is still no clear criterion for stationarity (a visual assessment of Prival's last graph, which is supposedly similar to BGS, is not a strict methodology). How to "restationalise" the process is another topic.

I just want to know how to check stationarity. And I need "residual" rows not for trading, but for checking systems (although no, I lie, I will check for professional suitability for trading as well).

 

Thank you, D500_Rised. Prival, do you have anything like that:

First and foremost is the<br / translate="no"> Dickey-Fuller test (or its extended version)- the ADF test
it's fairly well implemented in Eviews.

Structure shift tests are more difficult to implement
-Andrews-Zivot and
-Lumsdain-Pappel
 
olyakish:
NorthernWind:
olyakish:


Yes, there is a basis for this explanation. The main thing is that this "crowd effect" has a place in the graph.

The most interesting thing is that it can be used in trading ...

Yes, it can. And what's more, there are people who have been using these patterns for several years. With corrections of course.

 
Mathemat:

Here we are talking so much about stationarity and other clever things. But there is still no clear criterion for stationarity (a visual assessment of Prival's last graph, which is supposedly similar to BGS, is not a strict methodology). How to "restationalise" the process is another topic.

I just want to know how to check stationarity. And I need "residual" rows not for trading, but for checking systems (although no, I lie, I will check for professional suitability for trading as well).


Do not be upset. In fact, the concept of "stationary", whatever they say or write, has no clear formulation. And the practice of application... and the practice of application I tried to show just above.
 
Mathemat:

Thank you, D500_Rised. Prival, do you have anything like this:

First and foremost is the
Dickey-Fuller (or its extended version)-the ADF test
It's fairly well implemented in Eviews.

The more difficult to implement are the structural shift tests
-Andrews-Zivot and
-Lumsdain-Pappel
https://forum.mql4.com/ru/9321/page9
:о)
 
grasn:
to North Wind
I think I've seen something similar. What can I say, spectroanalysis as an auxiliary tool, who can object. It's a good idea to see if it's easier to use wizards or something like that instead of spectra.
Quite possibly, there have been several publications on the subject, and the issue has been worked out in detail on the forums. If I'm not mistaken, the entire adaptive window selection was reduced to the analysis of spectrum extremums based on the maximum entropy method. It seemed to work, and there was some sense in it since there are "plus" windows for MACD at any moment of time, you just need to find them. Of course it is possible to use MA instead of a spectrum, but depending on how the task is set, it may not work.

As for lowpass filters, I used to entertain myself by trying to predict the filtered signal using all known methods. In a way, of course, it works :o /

You can not just use MA, but a set of pre-calculated digital filters. It is possible that these very extrema do not "stray" much along the scale.

Besides, these peculiarities of candlesticks do not give us much hope for precision, so...

 

Guys, thinking about this topic. I have a simple problem:

Can someone explain to me in "two words" the meaning of the terms "Stationary BP" and "Non-stationary BP" ?

Due to the lack of this specifics, I can't be sure that I'm thinking about it correctly.

 
grasn писал (а): https://forum.mql4.com/ru/9321/page9
:о)

Grasn, on p. 10 there (I quote myself):

It turns out quite a tricky situation: in order to know if a process is stationary, one must first know its realistic model (here AR(1)). But it doesn't look like one. So the test doesn't seem applicable either.
Man, it's kind of a dead end. The definition of stationarity itself... is not the same, it's not strict. "For a process to be stationary, the m.o. etc. must be constant". I.e. the m.o. process must itself be stationary :))) Oiled...
 
Man, it's a dead end. The definition of stationarity itself is... something different, not rigorous. "For a process to be stationary, the m.o. etc. have to be constant". I.e. the m.o. process must itself be stationary :))) Buttery...

Where did you get that from? If m.o. is constant, then there is no m.o. process and no question of its stationarity, and as a result there is no buttery oil.

The definition that Prival gave is quite rigorous. What's wrong with it?
1...34567891011121314151617...32
New comment
Reason: