Random Flow Theory and FOREX - page 10
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
It's sort of proven in STO that it's of no use, as it's not a signal - much less a substance.
I don't agree about the point. This concept gives very much for the analysis of occurring processes. I can't imagine how the discipline of Radio Circuits and Signals would exist without this concept, and without it, there would be no televisions, no receivers, no cellular phones, and no computer on which you now work :-)
Can you decipher STO?
Are you sure the autoregressive model is first order and not second order? Although God knows what you are studying there ;-) Maybe you will write BP and how you do it. At least let me have a look at it :-)
That's the problem. The test is very specific and depends on the process model. A series is simply returns.
This is a rather tricky situation: to find out if a process is stationary, we first need to know its realistic model (here AR(1)). But returns doesn't look like that. So the test seems to be inapplicable.
Well Rosh has already answered it.
I am not saying that phase is useless. Are you referring to the phase velocity of the wave? Well then here it is: http://old.tspu.edu.ru/parfenov/kovo/ch4_6.htm. There are just a few phrases on the subject:
Phase velocity is a purely abstract mathematical concept, this velocity is not related to the movement in space of anything material.
The group velocity is related to the displacement in space of a perturbation of fixed amplitude; since the energy of a wave is related to its amplitude, the group velocity is the speed of propagation of energy in space.In general, phase velocity can exceed the speed of light (in the case of electromagnetic waves, for example, or De Broglie waves), while group velocity, in full agreement with the theory of relativity, is always less than or equal to the speed of light.
There is another experiment. We take a narrow beam of light, point it at the sky and follow the stars it hits. It is very easy to make the tip of this beam travel faster than the speed of light. You don't even have to point it at the sky, you can point it at some distant large object.
Phase in general is a very interesting thing, the only substance I know of that has a speed of propagation greater than the speed of light (this is an aside, if interested I can search for a mathematical proof of this phenomenon). Concerning positive values, most likely it is caused by two factors. The first is the features of the DFT try sin or cos. One will give + the other -. Second, where does such a huge number of signal components with the same phase (phase direction) come from. Try in my example to feed Y=t, i.e. a straight line equation, into the input. I think you will see, the subtraction of the 0th component of the spectrum in the indicator is not done quite correctly (I think you should Y-mu again :-) ). My attempts to use different Hemming and Butterworth windows didn't lead to anything, only made it worse (my teacher was right, application of windows is a tractor for energy). Therefore I've left the indicator in its present state. I don't remember, but somewhere I said, that this 0 component will still make us bleed :-).
One more thing about channels. In the course of mentioned topic (there is a link in "Stochastic Resonance" but it's a big problem to find it there :) I came to an idea of existence of some objects (entities) and channels are just their representation, maybe not the most successful one. Accordingly I understood the task as detecting and accompanying these entities. I must admit that a formal parallel with radar was just begging to be drawn, but it came to my mind only now :). In principle, if you are interested in "how it looked like" I can send you the ex4 indicator (about 60 Kb).
P.S. In fact it's not important for me whether we communicate on "you" or "you", but if the person goes on "you" and then comes back to "you" involuntarily there is an idea - whether I have offended him unintentionally?
Here are more useful articles: http://www.nsu.ru/ef/tsy/ecmr/study.htm
Maybe someone will find them useful in terms of development.
I really don't care if we are speaking on a first name or a second name, but if a person switches to a first name and then comes back to a first name one involuntarily thinks - didn't I offend him unintentionally?
Not at all, on the contrary, I'm very grateful to you for your help. I don't care what they call it either, "even with a pot, just don't put it in the oven :-)". As for Matkad, put it in, spend some time studying it. I have Help-books, books, examples, all in electronic form, I can send you. By and large, it is not a programming language, there is nothing to program. As you see the formula in the book, you write it, it calculates everything by itself. And it outputs it in the form of graphs.
As promised, I am now writing down the purpose and route of the research. That's a lot to take in. I'll post it soon.
As promised, I will give you the direction to go with this phrase
The goal without a way to achieve it - a mirage, the path without the goal - a road to nowhere.
The purpose of the research - as in all of us, building an automatic trading system that brings profit.
The way - searching for patterns of behavior of the curve of exchange rates, allowing to use them when building an automated trading system.
Stages of work.
1. Proposing a hypothesis and searching for a mathematical apparatus which allows studying this curve (output).
2. Searching for evidence of the postulates of the Theory of Random Flows and FOREX (issue of Common Sense + Pvr42 indicator)
3. Reducing the curve to a form suitable for analysis (obtaining the ergodic process). Partially performed on a 1-week sample.
4. study of the obtained time series (TR) by ACF at different currencies and time intervals (stopped here, because we need an indicator).
5. While preserving the type of ACF, searching for the function approximating it (it is better to use an analytical function). Collection of statistics of ACF parameters for various currencies and important timeframes (calm market, news release, etc.).
6. Constructing a model of curve "behavior" on the basis of studies and checking its adequacy (obtaining the F transition matrix, see 1.p.).
7. When we obtain adequacy of about 70%, construction of a multivariate Kalman filter using the obtained model(s), each filter is adjusted to its own parameters (calm market, news release, etc.). (Explanation: the Kalman filter allows to get a forecast !!! of the movement by ISC). Inconsistency - mismatch of the filter output and the incoming information + operation of another Kalman filter (working in parallel) will give us a sign of change of "behavior" of the market (change of the model incorporated in the filter) - the decision point.
8. Study of distribution laws (PD) of discrepancy at the filters output, for construction of a solving rule of market entry. IHMO the best two threshold Wald detector. (The logic of operation is based on Yes-No-No, and accumulation of statistics before making a Yes or No decision).
9. Setting up the ATC, testing all blocks and procedures. Taking measures to ensure smooth operation and manageability + taking measures to protect the algorithm.
Computational load will be not just big, but huge. Some optimal solutions can never be obtained, as they require infinite computing resources (damn mathematicians proved it :-(), but every cloud has a silver lining. Since there are infinitely many variants of Kalman filter system and decision-making options, there is enough space for all of them. These infinite procedures have to be trimmed, and this is more an art than pure mathematics.
I appeal to all.
1. I have books explaining many concepts and procedures in electronic form. I will give them to anyone who asks.
2. The only thing that is valuable in this world is TIME, and unfortunately I spend it on writing such posts. If you will not help me, then I have only one way out. Hire a programmer who is good at MQL4, perhaps he will be satisfied with a part of beggarly officer's salary or do it myself. I have programmed in many languages, starting from Focal, Basic, Fortran, Assembler, ..... Matlab, Matcad. I chose the latter because it is the most efficient (time-saving) for testing various hypotheses and conjectures. Usually I used to give the worked out algorithms in this language to the software engineers in research laboratories, which I managed. Now, unfortunately, I have a 70-year-old grandmother in this position.
3. Don't think that there is not much you can do to help. I will try again to show it by example, at the same time I would like to express a BIG thank you to these people.
- Mathemat not so long ago asked about calculating ACF and asked to do a selection of books for him. That's why I wrote these 9 articles (and this forum thread appeared in general) as I remembered that about 12 years ago I did the same analysis. Many of those materials have not remained, only scraps of any articles, research work, programs and it is necessary to restore all from memory. But I'll manage.
- lna01 helped me a lot to write the indicator I mentioned above, and its questions make you think. Trying to answer them gives a direction for new researches that may lead to creation of new, interesting indicators. The results of his work, I think, can already be used as one of the inputs of a neural network. (I think it's better not to use the"Non-standard automatic trading" approach with going from -100 to +100, maybe something will turn out well, but to make some sense reading the theory of recognition, for example, build a Voronov chart and, most importantly, feed different indicators (signs of recognition) to input values, one of which may be energy).
- Neutron his references, scripts allow you to look at what we are doing from a different angle to enrich yourself with new knowledge, it is also important. The only thing is there is not enough time for everything.
All we communicate we enrich ourselves with new knowledge, ideas, time will show what comes out of it. And whether I will be able to go the way of the 9 points, too, I do not know. But I will try very hard.
P.S. As the creator of this branch, I beg you very much, try at least less flooding, you know yourself as hard as 100-200 pages to find that grain of really important information. Do not introduce "philosophical" concepts, ie those that can not be written language mathematics. If you do not understand something, ask. Putting the right questions is 2/3 of the answer, although there are questions that even 100 wise men can not answer. I'm not God and do not know all the answers, but all their knowledge with pleasure to share with you, most importantly, that would be enough TIME.
S. Respect to all. Privalov
If you want help, just write Privalov, I want to help. I will tell you what you need to do, and I will try to make it interesting and within your power.
Cool!
Prival, don't mind giving a theoretical justification for the applicability of this method of analysis to Time Series like exchange rates. 2-3 points of general and short considerations will be enough. If it's boring - don't answer, just don't send it to the beginning of the thread - I was there and didn't understand it :-(
Let me try this again with an example.
The main thing is to understand this formula
A simple example, suppose you know that time series (RT) obeys the law y(t)=a+b*t. Knowing a and b and the state in which I am at time t, it is easy to calculate y. This is what is written in this formula, only in matrix form + some subtleties, more on them later.
I'll write it as L(k) is a state we are in now (in the example it's t), next state L(k+1) depends on matrix F (called transition matrix) in the example it's coefficients a and b.
Now for a nuance. Markov process. According to this theory, the transition from the state L(k) to the state L(k+1) does not depend on the state L(k-1), i.e. the rate was the same yesterday, an hour ago, a minute ago. The main thing is the exchange rate L(k). What it will be at L(k+1) is determined by this damn (I cannot find another word ;-)) matrix Ф.
Most likely, F will have different parameters for calm market,news release, etc. The main thing is that it should have such a view as I drew earlier on the real sample. Let's put it into Kalman filter (there are several of them, each adjusted to its own model with different parameters of this curve). Kalman filter gives us a forecast by OLS what price should be at L(k+1) point of time. When price arrives, we compare it with the forecast and the filter with the smaller difference will ring (jargon). Like in the receiver, you twist the knob and hit a station (trend) and hold it. Missed a station, turn the knob until it rings aha news, worked off the news, gone again went to turn further. So there you go. But it's more complicated and in formulas.
That is why it is so important to find the F-matrix.
Now ACF, in our example it is always equal to 1. So it's easy. But we can determine this transient matrix by looking at the ACF.
Edit. This is exactly the reason why I asked the MQL developers for a paralleling library. To parallelize these Kalman filters (not to twist the knob ;-). So that if you don't have enough power you can build a cluster.