Market phenomena - page 7

New comment
 
Avals:

I checked - there was nothing like that :) That's why I'm asking how you built, rounded, etc.

so it's an optical illusion :o) You should check it, just in case. After all, all the numbers are there, including the classification.

 
RomanS:

What is the quotation process?

if it were to be reliably known ...
 
Farnsworth:

If it were known for certain ...

Are you still developing strategic forecasting systems, it would be interesting to see the forecast?
 
Vitya:

Are you still developing strategic forecasting systems, it would be interesting to see the forecast?

And that's it, the contest is closed. (Let me remind you the participants: Yusuf Khoja, this Matemat and Volnoviki (there are a lot of them abroad)).

Leader: Yusuf Khoja was withdrawn - they found doping (a lot of weed there...) Advisor - written off...

2nd place (went smoothly into first place) (the more obscure the better - who's going to argue?)

3rd place (the topicstarter - didn't get off the mark...(for objective reasons))

 
Farnsworth:

There are several areas of research that I have been working on for a long time. One of them is the bold assumption that the increments of a quotation process are essentially a superposition of several, simpler processes. This "superposition" could be a sum, a product or a more complex transformation.

Why? If only because the quoting process is supercomplex, but not random at all (they are different classes of processes and can be distinguished if desired). This is a separate topic for conversations over a cup of some liquid, but by the way, there is proof.

The phenomenon that I want to present, perhaps, is known to someone, and maybe not, or known not to all. Anyway, I haven't seen its mentioning anywhere. Let's take EURUSD M15 (Alpari data for approximately 10 years) and look at its increments.

Now let's look at the histogram, but usually they are looked at like this

or like this:

I looked for this manifestation of superposition in every way, applying the most perverted methods and, ... and it appeared, directly, in plain sight, or rather one of manifestations. And it suggested the place of search - the so-called "subtle structures", if one can say so, and that's what I found.

Now look closely, first argument of histogram function is number of intervals, by which frequency of events will be counted (graph is limited, i.e. very long tails), graph appearance is changed:

h=100

h=200

h=300

h=400

h=500 (something appears)

h=166

h=700

Further increasing h already merges everything, nothing will be visible.

You can clearly see that a small process sits inside the big one; I gave them names, processes "alpha" and "omega". I've named them "alpha" and "omega". Now I need to classify them applying scientific groping method I've got the following matrix for classification of two processes:

Column designations:

  • first column - number of system state
  • Second column - beginning of interval for the class
  • The third column - end of price interval for the class
  • fourth column - type of the process

Now we need to go through the whole time series of M15 increments and collect these two processes, which I am doing. To clarify, "assemble" means to add up all the increments taken over these intervals for each process class. Clearly there will be omissions, but I don't take them into account now, i.e. if zero is added for the opposite class event.

ALPHA process (the whole process and a fragment of its increments):


OMEGA process (the whole process and a fragment of its increments):

Here they are, the bulls and the bears.:о) But no, I don't believe in those animals, I think this zoo is bigger in Forex, it's a jungle there. Now we can move on to the market model. It fits well, well ... almost well to the basic model, which I use - stochastic systems with a random structure (described briefly here https://www.mql5.com/ru/forum/129406/page15).

It turns out that there are almost (!!!) two linear processes, between which there is switching, i.e. there is another possibility to describe an adequate model, well ... theoretical at least :o). You can calculate the transition matrix from state to state. And it may seem - this is "it", no, this is not "it" yet. When "it" comes - I will tell :o) there are really complexities, processes are not linear, here is an increase in a fragment:


Or it needs to be more precise (I've cheated), but there is a lot of noise in it, bad linear correlation, transitions are not clear yet, but there seems to be no "Markovism", i.e. there is a dependence and you can't find it.

In general, colleagues, it is possible to discuss the philosophy, theory and practice. By the way, the phenomenon works on all relatively small t.frames.

I can support the author in the fact that I myself have got similar density functions, i.e. with alternating peaks and dips over the whole range. I even wanted to do something with it, but forgot about it.

I would like to notice, that I made a histogram not on first differences, but on their moving averages. I also get a "scallop", if I'm not mistaken...

 
Vitya:

Are you still developing strategic forecasting systems, it would be interesting to see the forecast?


So far, theoretically in the sense that I'm trying to find a way to refine the forecast. Of course it is good that the model is accurate enough (correlation of first lags of model error is less than 0.03), but time "kills" the model, every lag brings new opportunities of price movement and it is very difficult to catch a confident strong movement for a couple of months. But I am thinking about it. It is like if you do it for a long time, you will get something. I think there will be predictions, just let me think about it :o)

 
Tantrik:

And that's it, the contest is closed. (Let me remind you the participants: Yusuf Khoja, this Matemat and Volnoviki (there are a lot of them abroad)).

Leader: Yusuf Khoja was withdrawn - they found doping (a lot of weed there...) Advisor - written off...

2nd place (went smoothly into first place) (the more obscure the better - who's going to argue?)

3rd place (the topicstarter - did not leave the start ... (for objective reasons))


"we'll see who's who" (C)

Here will expand consciousness to the level of a hodja and that's it, I'll take first place :o)

 
alexeymosc:

I can support the author in that I myself have had similar density functions, i.e., with alternating peaks and dips throughout the range. I even wanted to use it in some way, but forgot about it.

I would like to notice, that I made a histogram not on first differences, but on their moving averages. It also turns out "scallop", if I'm not mistaken...


Thanks for the encouragement. It's really hard to use this phenomenon, but let's see if we can find a way.

 
Farnsworth:


Thank you for your support. It is indeed difficult to exploit this phenomenon, but let's think about it and see if there is a way.

I agree. Maybe the phenomenon itself will be found.
 
paukas:
It didn't used to be, but now it is! Phenomenon :))


Pakukas, the phenomenon is there, the reasons why it's not seen are very simple:

  • If you build a mma with a minimum pitch, you usually cram hundreds of thousands of indicators into a small window and you just can't see anything.
  • They often take a big step, and everything is aggregated beyond recognition.

but what if it's there? Paint your beard a pungent yellow? All right, just repaint the beard on your avatar. And if it's not there - I'll leave the forum and I certainly won't bother with your inability to use TA and VA anymore. Deal? :о)))

1234567891011121314...75
New comment