Stochastic resonance - page 7

New comment
 
Yurixx:

Do you guys really think you can build a fundamental theory here?

We think (I think the majority would agree) that it is necessary to think, and if this theory does not lead us to the grail, then forget about it, remember the Indian prince, or maybe not even the Indian prince, who never solved the problem, but found great answers to questions close to the problem set. Perhaps someone will find something of value in the discussion of this question. It is an unseemly position to convince others of their incompetence and futility of their efforts. It seems to me that fundamental theories do not fall from the sky, they arise in such discussions as well.
 

2 Yurixx:

The result can, and even should, imho, be simple (in use). But the road to it can be very convoluted and varied.

 
lna01:

2 Yurixx:

The result can and even, imho, should be simple (in use). But the road to it can be very convoluted and varied.


Absolutely, and if the high assembly wishes to engage in stochastic resonance, then I wish you well.

However, imho, the stochastic resonance model itself is too weakly suited to the market, as you actually agreed.

2 AAB

It is an unseemly position to convince others of their incompetence and the worthlessness of their efforts .

Dear Sir, you have something wrong with your eyes. Cite the quote you so "originally" interpreted.

 

to Ina01

Of course I'm the one Candid, who else can be found in such a thread :). So hello back! :)

Hi! I suspected as much right away. But why do not call himself Candid, I think with such a nickname there is no one.

So the standard model for SR consists of the dynamic part (read potential), cyclic signal and noise. So if you insist on SR there is no getting away from them. Of course we have to choose (to start with) the simplest but adequate model. IMHO, if we are interested in direction it should have three stable states - one current, one below and one above. They don't have to be called potential pits :), but you can call them so. The most controversial part (imho) is the cyclic signal. And I have no objection to a regular time structure - we're not even talking about sessions, weekends, quarterly reports, elections etc. Sometimes I think that "news" was invented by someone extremely clever (or cunning :) - exactly to regularize signals and consequently, to simplify calculations in a model unknown to us :). The only problem is that the amplitude and sign of the signals are, I suspect, quite close to random (at least with our degree of awareness)."


For this particular case I just want to start with the search for patterns related to noise, and not to build models and argue with a hoarse voice for the presence or absence of cycles, for example. I've been working by the principle: if you don't know something - ask the market.

The formula for noise intensity ... Can you define it (intensity)? That's not an excuse, the formula is only possible for a specific value (say, the value given in the model).

In all read sources they write about noise intensity as some completely understandable and well known quantity. But they don't write clearly what it is. I even developed an inferiority complex about it. :о)

to Yurixx

If, as Sergei says, you have to create a scientific research institute, why do you need it at all? Or are you going alone to compete with large and well-funded groups that serve foundations?


Sergey is just joking :o)))) In general, I had to involve professional mathematicians in developing the model.

Look at the championship leader's expert. It gives 80% profit a week, and it is based on oscillators, good MM and non-trivial logic. And it was made by one person. I do not know how it will work further, but the result of these two weeks is almost ideal for a single person.

Yuri, my Expert Advisor, conscious on the basis of all the same fundamental model, shows much better results than even the sum of the leaders of the first top. And this is not bragging. I tested the forecast on every bar and it showed error less than 1% for major quotes and different DCs. Here is an example of how the model works, so far in MathCAD (works at least on hours):

The red square is the expected reversal zone, consisting of calculated level that price will occupy and estimated probability of reversal in this zone. Probably, you will ask why I'm not in the Championship? The answer is very simple, I will not be able to fulfil the condition of the five-minute test, and I do not need the Championship. Besides, I'm transferring the code, and the main snag is the problems with calculation of the moose levels. There's quite a lot of error (about 60%).

So why did you enter wavelets and as far as I understand from a neighboring branch want to involve NS in wavelets messing around?

All my life I have been fond of global theories. From the roots, to the axiomatics, to the specific calculations of the values used in practice. But the more I deal with Forex, the less I'm attracted to it. Why this global fundamentalism (or fundamental globalism :-) ? Do you guys really think that you can build a fundamental theory here ? Then at least try to outline the range of phenomena that underlie forex, identify the basic objects and their interactions that define its nature.

You seem to have been attacked by pessimism - the basis for the development of more complex illnesses. Good advice, a good mug of ale, in the company of friends, can easily cure this ailment. I remember you "young", so to speak, when I first met you on the forum. That was when you had a very different opinion. :о)))

Can we discuss how such a thing can even be ? :-))

Better let's discuss how to count noise intensity!!! I assure you, it's much more useful than just chattering about the championship.

PS: and if we analyse the championships, it seems that none of the participants have ever held stable prizes throughout the championships (and everyone has sort of participated). Or am I wrong?

to Mathemat

I do not understand something with potential pits, I've obviously missed something. Where can I see it? In the Niroba branch on metaquotes?

That's right, you can find some good thoughts on potential application there.



Big PS: if no one wants to pursue this avenue, I didn't really want to. :о))) Tell me how this XXXXXXX (crossed out and again crossed out) noise intensity figure counts and I'll get off everyone's back. :о))) I give you my honest, noble word - I'll get off.

But I assure you, it is an excellent model, including for the market, it just needs a little bit of refinement, much better than the mashka that Yuri admired :o)

 
grasn:

Hi! I suspected as much from the start. But why didn't you call yourself Candid, there's no one with that nickname.

I did call myself Candid, but it turned out to be a username instead of a name :)
 
grasn:
Look at the expert of the championship leader. It yields 80% per week and is based on oscillators, good MM and non-trivial logic. And it was made by one person. I do not know how it will work further, but the result of these two weeks is almost ideal for a single person.

Yuri, my Expert Advisor, conscious on the basis of all the same fundamental model, shows much better results than even the sum of the leaders of the first top. And this is not bragging. I tested the forecast on every bar and it showed error less than 1% for major quotes and different DCs. Here is an example of how the model works, while in MathCAD (works at least for hours):Identify the main objects and their interaction that define its nature.

I remember you "young", so to speak, when I first met you at the forum. That was when you had a completely different opinion. :о)))

I dare to disagree with you. Your model is as much a phenomenology as any other. Though, one must assume, quite complex. And stochastic resonance is also phenomenology. A fundamental model must proceed from the definition of basic interaction objects, types of these interactions and their laws. Take any physics theory: an object, its fundamental properties, parameters that describe these properties quantitatively, laws of interaction expressed in the form of equations. Where is it on the market? I have never seen anyone even try to formulate what is an interacting object on the market, not to mention everything else. I therefore immediately reject the fundamental theory option as currently impossible. And so the use of stochastic resonance is an attempt to build a phenomenological theory, but based on a specific model. The usefulness of which I questioned.

And in general, I want to give you credit. 1% prediction error on every bar is an impressive result. Taking part in the championship is a toy. If you put your brainchild, on which you have spent a lot of time and effort, in the championship, I would not understand you.

And you're wrong about pessimism. Since I'm in the business of building a phenomenological model myself, that means I believe in it. My doubts about SR are simply due to the fact that the more complex the model, the more time it takes to realize that it doesn't work. And time is something we don't have much of. So I haven't changed my mind yet. But it's nice that someone still remembers me young. :-))

 

to Candid

I did call myself Candid, but it turned out to be a login instead of a name :)

It happens. :о))) In general, of course, dear Metakvotes it would be desirable to make a portal with one input, but the abundance of separate, unconnected sites about the same thing is not very convenient. If I write "to Candid", it will be understandable, because I'm already used to it? :о))))

to Yurixx

I dare to disagree with you. Your model is as much a phenomenology as any other. Although I have to assume it's quite complex. And stochastic resonance is also phenomenology. A fundamental model must proceed from the definition of basic interaction objects, types of these interactions and their laws. Take any physics theory: an object, its fundamental properties, parameters that describe these properties quantitatively, laws of interaction expressed in the form of equations. Where is it on the market? I have never seen anyone even try to formulate what is an interacting object in the market, let alone everything else.

You may be right and the model is not fundamental at all. It is based on the stochastic model (mainly by Shiryaev), fractal theory (not those fractals that can be found as an indicator), some postulates of the theory of random processes (the same Shiryaev) and Elliott theory (I keep forgetting where to put some letters). This allowed us to abstract away from the physics of the process. But all the other components are present: objects, relations between them, and dependencies. Formulas derived empirically, based on the results of Data Mining and statistical analysis.

PS: I'm not upset about it, let it be phenomenological. Just like that, you killed my pride with a few sentences, by the way, thank you :o)))

That's why I immediately reject the variant of fundamental theory as impossible for today. And so the use of stochastic resonance is an attempt to build a phenomenological theory, but based on a specific model. The usefulness of which I have questioned.

In conversations with Candid, I described what so "grabbed" me about this approach. I now ask you a direct question, how do you calculate the noise intensity? If you've seen it somewhere, please tell me, I can't find it anywhere, I have a feeling that the special services have cleaned up everything :o)

In general, I want to give you credit. 1% prediction error on every bar is an impressive result. If you were to put your brainchild, on which you have spent a lot of time and effort, in the championship, I would not understand you.

Thank you, I have spent a total of about five years on it, if you count from the first ideas in this area. The System will not be at the championship, it has a different target function.

And about the pessimism, you are wrong. Since I'm in the business of building a phenomenological model myself, it means I believe in it. My doubts about SR are simply because the more complex the model, the more time it takes to figure out that it doesn't work. And time is something we don't have much of. So I haven't changed my mind yet. But it's nice that someone still remembers me young. :-))

It's good that I was wrong, pessimism is a bad companion in Forex. :о)))

 

Yes, I was going to write about noise, but somehow it fell out.

In the physics analogy, the intensity of noise should be related to its energy in direct proportion. The energy of an oscillating process in physics is proportional to the product of the square of amplitude and frequency. You already know the frequency and spectrum. In Forex the amplitude is the range of price fluctuations, or (sorry, old man) - volatility. :-))

In principle, the square of the amplitude can also be replaced by the first or other degree. If there is a specific model, then this indicator can be made a parameter for now, and its exact value can be determined experimentally by comparing the experimental data with calculations on the model.

Suitable ?

 

to Yurixx

The energy of an oscillatory process in physics is proportional to the product of the square of amplitude and frequency.

This is just a homogeneous quadratic energy function for oscillators (I mean the classical ones that are written about in physics textbooks). It might work, I guess I would have to do a direct Fourier transform for the signal and slip the resulting amplitude and frequency to calculate energy using this formula. Suppose I calculate this and this model could theoretically be taken as the first draft. I can also calculate the power spectrum according to the DSP. But how do I calculate the noise intensity in these cases?

If by analogy with physics the noise intensity should be related to its energy by direct proportionality

Somehow they are related. It seems logical, the more energy the system has, the more noise it should make. Only it is not clear to me how to calculate the proportionality coefficient, since it is unlikely to be equal to unity? I can't figure it out.

 

The concept of "noise" does not exist without "useful signal". i.e. you must first define what a "useful signal" is, noise will be found automatically. A useful signal cannot be defined without a formal description. Noise is found automatically as a difference between a signal and a "useful signal". If this difference cannot be described with the current knowledge of the process, it is called "chaos".

In terms of trading - the processes on small frames influence the formation of the more global ones if there are preconditions for that. For example, a breakdown on the minutes will start an avalanche which will form a trend on the daily chart, but if there are preconditions for that. The potential is accumulated and its realization will start due to "chaotic" moves. Chaotic, of course, only from the point of view of a particular observer (his information about the process). Potential in the market can be thought of as the pent-up desire/need for a significant part of the capital to make certain transactions. Sentiment, one of the names of this potential.

1234567891011121314...38
New comment
Reason: