Stochastic resonance - page 10

 
Mathemat:

p.d.f. - probability distribution function, a probability density function. "Black swan" is Taleb's term from his Fooled by Randomness. It is a rare but crushing event whose frequency in the market is much higher than what it should have been under the "normal hypothesis" of the returns distribution. References:

http://stock01.narod.ru/ - both Peters' books are there at the very end. I will add a link to Taleb here a bit later.

ok, thanks, and taleb found www.internettrading.ru/bibl/pdf/taleb.pdf
 
grasn:
Avals:

Vinin is right:

http://....

And right about what? He wrote "Sorry, no answer". This is a reference to calculating signal-to-noise ratio (I've already programmed it) and has nothing to do with noise intensity calculation. But what is needed is noise intensity, which is what stochastic resonance theory stresses and makes the distinction between the two.


The crux of SR is that the signal is unable to cross threshold (U) without noise (D). In equations of SR the U/D ratio must be dimensionless, so if threshold is in pips, noise must be in pips. In general threshold is overcome due to noise amplitude while noise frequency must be much bigger than frequency of useful signal but it has no impact on the final result if our task is to recover useful signal. What's the problem anyway? :)

 
AAB писал (а): and Taleb found www.internettrading.ru/bibl/pdf/taleb.pdf
The link should be corrected, AAB. If you click on it, it's wrong (see link properties). And it doesn't open...
 
Mathemat:
AAB wrote: and Taleb found www.internettrading.ru/bibl/pdf/taleb.pdf
The link should be corrected, AAB. If you click on it, it's wrong (see link properties). And it doesn't open... Is it "Fooled by Randomness"?
Yes, there really isn't a file, Google gave a search for Taleb, in the First position on the link http://www.google.com/url?sa=t&ct=res&cd=1&url=http%3A%2F%2Fwww.internettrading. ru%2Fbibl%2Fpdf%2Ftaleb.pdf&ei=v10TR9iBIJeM0QT1ifmCCw&usg=AFQjCNE0FLo1HJTomEX35i9m95TD4GMy8g&sig2=sPU6EDQHDLcJ19lY7PivnA
I pulled the taleb.pdf file, 1.05m, and didn't want to give the link with google search tags (you can see how "dirty" it is), I copied the link from the search engine text, what a bummer, sorry .

Yes it's not all right, it's that site fluctuates (makes noise:), I once will appear once no, search led to "Fooled by chance"
 

2 grasn

A little more on the physical sense. If you take not the square of the amplitude but, as I said before, a linear relationship, then the value P=2A*f (where A is the amplitude of price fluctuations and f is the frequency) is a prefit, which can be obtained within the spread for one complete swing of that frequency. Here is the coefficient that arises naturally. And at the same time the target function is associated with the intensity and can be used to determine the maximum return that can be extracted in the market in some time T and hence determine the efficiency of one's strategy by comparing its yield to this value. In this case the maximal return turns out to be linearly related to the energy of the market (=total intensity of all components). So this option is even better than E=f*A^2.

 

to Yurixx

Avals Completely rightly stressed the presence of signal and noise in the same stream. How to separate them ? There are only two options: either define signal or noise. To me personally the first option seems more logical. We go into forex hoping to predict the trend (even if it is a flat). If the prediction is correct, then we can make profit from it. However, we are not trying to "decipher" all of the information contained in the price chart. That's why when we define trends we want to identify (or determine their presence) in the data flow, we will get conditionally three signals (up, down & flat) which are interesting for us. Everything else is noise.

Quite right, and I understand it very well, starting from the second page :o). But speaking of noise, I can say for sure, that it will be much and of various types. :о)

It follows from all this that it makes no sense to consider noise intensity as power (i.e. in J/sec) either. So the only thing left is to interpret the intensity in this case as the energy of vibration of the price. The sense of it is clear - we cannot divide Forex into parts, components, units, sequential processes etc. The whole process exists as a whole. In this sense the price data stream reminds me personally of a signal received from the space: it is unknown who generates it, what information it carries, in what language, what is the encryption key, etc. But we can calculate the energy of this signal. Naturally, with limitations imposed by Heisenberg's uncertainty principle.

Perhaps, but as an indirect estimate, but it seems to me that it is still a relation. Anyway, whatever, I will collect both energy and my version of noise intensity estimation and maybe a couple more.

Candid's remark about perceived range is great. Everyone has their own speculative ideas. One wants to play on the minutes, the other on the days. It is clear that for them the perception of trends will be quite different. And so the same what the day trader sees as a trend, the one who plays on the days will see as noise. So, imho, one should not take the word signal in an absolute sense. It would be more correct to define one's own interest and filter it out of the data stream.

I've written before and continue to yell about it now - you cannot let trading (changing us), technical analysis together with fundamental analysis anywhere near this research. All that should be left (and which I have taken care of) is 'scale' as a parameter and nothing more. Let them get closer - we won't find anything, but this way there is a chance, albeit a small one.

One last thing. Energy is directly proportional to the product of the square of amplitude by frequency. But the coefficient of proportionality, believe me, does not play a role. It's only needed to keep the dimensionality.

And again it is quite possible, I just do not remember, but it seems to me that for physical oscillators, (whose nature is taken abstract by the same wave theory), it turns out, no matter how twist/extract it is always 2. But I'll trust the physicist, let it be a parameter as well.

So don't suffer, don't poison yourselves with beer, but brave the battle. :-))

How can you write about beer like that? It is the Earth's drink, with all its strength and wisdom imbibed in it! :о)))) And I'm not poisoning myself at all, but calmly planning an experiment. :о)

A bit more about the physical sense. If you take not the square of amplitude, but, as I said before, a linear dependence, then the value P=2A*f (where A is the amplitude of price fluctuations, and f is the frequency) is a prefit, which can be obtained for one full fluctuation of that frequency, with the accuracy of spread. Here is the coefficient that arises naturally. And at the same time the target function is associated with the intensity and can be used to determine the maximum return that can be extracted in the market in some time T and hence determine the efficiency of one's strategy by comparing its yield to this value. In this case the maximal return turns out to be linearly related to the energy of the market (=total intensity of all components). So this variant is even better than E=f*A^2.


I don't really understand which 2A*f we are talking about? Is it the sum of them after representing the price as a harmonic pile, or the frequency at which A is maximal?

to Avals

The point of SR is that the signal is not able to pass the threshold (U) without noise (D). In SR formulas, the ratio U/D must be dimensionless. Therefore if threshold is in pips, noise must also be in pips. In general threshold is overcome due to noise amplitude while noise frequency must be much bigger than frequency of useful signal but it has no impact on the final result if our task is to recover useful signal. What's the problem anyway? :)

Right, it all depends on what the task is. I wrote many times, what I want to do - namely, just to collect statistics, if I may say so, about noise influence on flatness stability and try to find regularities. In other words, I'm not building five-storey formulas symbolising the market right now, and I think it's absurd. What the SR is all about and probably read the same publications. In this regard, just forget about SR for a while. There is only noise and specifically you need to find noise intensity. The term exists irrespective of SR.

 

Не очень понял, о какой именно 2A*f идет речь? О их сумме после представления цены в виде кучи гармони, или о частоте, на которой максимальна A?

It's about a variant of interpreting the physical sense. Since it makes sense :-), you can trade on every harmonic, or you can only trade on the main one. It all depends on how much money you need.

 
Avals:
What's the challenge anyway? :)


I picked up a picture, it seems to show quite well steady states (not only horizontal flats, but also trends) and jumps between them. From it, imho, it is clear that the aims of the jumps are predictable to a large extent. I'd like to see if the timing and direction can somehow be predicted :). But, again, I think that stochastic resonance as a mechanism of transitions has nothing to do here.

 

here's another one for ~5 pennies

here about Power Spectral Density...

https://en.wikipedia.org/wiki/Power_spectral_density

or

Energy per symbol /per noise of power spectral density(Es/N0),

https://en.wikipedia.org/wiki/Eb/N0

 

2 grasn:

Regarding my previous post with the picture: I wonder how much our questioning is the same at this level?