_Market description - page 29
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
Trading real on forex is like having sex with teenagers:
- everybody thinks about it;
- everybody talks about it;
- everybody thinks their neighbor's doing it;
- almost nobody does it;
- whoever is doing it is doing it badly;
- everybody thinks they'll do it better next time;
- nobody takes safety precautions;
- anyone is ashamed to admit they don't know something;
- if someone succeeds at something, there is always a lot of fuss about it.
You're a beauty! >> Did you come up with that yourself?
You know, I often find it funny when people try to explain to me what has long been known and do not understand what I am talking about (I understand them perfectly well). I have a monograph devoted to spectral processing methods, scientific works and research in this field.And you are trying to make a fool of me. I am a fool and you are right...
Z.Y. Your delusions are more true than mine :-))
So ... ? What does it mean that you have books? That you are never wrong? Does having books on the wrong application of the Fourier method make you infallible and put you beyond criticism?
LProgrammer has clearly explained to you where you are wrong.
So... ? What does it follow from the fact that you have books? That you are never wrong? Does having books on the misapplication of the Fourier method make you infallible and put you beyond criticism?
LProgrammer has clearly explained to you where you are wrong.
If you can't see where he's fudged the maps, too bad. He suggests that the function k*x+b should be decomposed into a Fourier series. This function has an infinite spectrum, and according to Kotelnikov's theorem you need a sampling rate of 2 times larger.) I've been telling him about this theorem for already 5 pages, he obviously does not know it, he has only heard about the frequency of Neukvist ..... And the consequence is direct from the theorem, you can decompose only a function that has a limited spectrum. As for sinusoids - any curve can be represented as a sum of sinusoids. And the Fourier transform is just a transition from one coordinate system to another. You had a function that depends on time (amplitude - time, what we all see on the screen), did the Fourier transform and got this function in amplitude-frequency coordinates. And that's it. Someone wants to do analysis in the time domain, build mash-ups, RSI, FIR, BIR filters, etc. And someone wants to look at the market from the other side (from the frequency domain) and try to analyze it by building the same filters, it's easier there (to build a filter).
And he is wrong to pick on spectral analysis. If he does not want to work with a spectrum, let him not do it. Talking about the fact that the task is not to build a spectrum, as he put it "shitty", but to continue the curve into the future is necessary. Yes we do, we really do. We all do. But to think that PF will solve the problem of nonstationarity in time domain is fundamentally wrong. In time domain it is unsteady; in frequency domain spectrum is floating. If the spectrum didn't float, in the time domain it would be a clear periodic function.
Z.S. Everyone can be wrong, and so can I. I've made so many mistakes and made so many mistakes that my forehead hurts (I've spent a month looking for an error, I've translated the program from MathCada to MQL, I found - number Pi was not set precisely and the error 'Pi' gradually accumulated). I have plenty of them in electronic form too.
If you can't see where he's fudged the cards, too bad. He suggests that the function k*x+b should be decomposed into a Fourier series. This function has infinite spectrum, while according to Kotelnikov's theorem it is necessary to have sampling frequency 2 times bigger). I've been telling him about this theorem for already 5 pages, he obviously does not know it, he has only heard about the frequency of Neukvist ..... And the consequence is direct from the theorem, you can decompose only a function that has a limited spectrum. As for sinusoids - any curve can be represented as a sum of sinusoids. And the Fourier transform is just a transition from one coordinate system to another. You had a function that depends on time (amplitude - time, what we all see on the screen), did the Fourier transform and got this function in amplitude-frequency coordinates. And that's it. Someone wants to do analysis in the time domain, build mash-ups, RSI, FIR, BIR filters, etc. And someone wants to look at the market from the other side (from the frequency domain) and try to analyze it by building the same filters, it's easier there (to build a filter).
And he is wrong to pick on spectral analysis. If he does not want to work with a spectrum, let him not do it. Talking about the fact that the task is not to build a spectrum, as he put it "shitty", but to continue the curve into the future is necessary. Yes we do, we really do. We all do. But to think that PF will solve the problem of nonstationarity in time domain is fundamentally wrong. In time domain it is unsteady; in frequency domain spectrum is floating. If the spectrum didn't float, in the time domain it would be a clear periodic function.
Z.I. Everyone can be wrong, and so can I. Don't worry, I've made so many mistakes and made so many mistakes that my forehead hurts (I've spent a month looking for an error, translated the program from MathCada to MQL, found - number Pi was not set precisely and error 'Pi' was gradually accumulated). I have a lot of books I can give to all comers, I really do have a lot of them in electronic form too.
I will repeat what LProgrammer told you, what decent modern mathematicians know, and what Lagrange and his fellows said in their time, when they AGAINST wide application of the Fourier method, and I will show you your fallacies, which are PROSECUTION of important conditions. This is nothing new, it is described in Fink's books, in the works of Academician Ageyev, in popular form it was even somewhere on "Computer":
Kotelnikov's theorem is not a theorem because it DOES NOT define the concept of "spectrum". What is a "spectrum" in that theorem? It is in that theorem nothing more than FURIER DIVISION. So the theorem is not a theorem at all, it is just a tautology. I am not the one calling it a tautology. It is stated even in the old books of Academician Harkiewicz.
You are PROPERly slipping into fallacies like "As for sinusoids - any curve can be represented as a sum of sinusoids." Yes, it is true, only you do not say that "as an infinite sum of sinusoids". And Kotelnikov's theorem immediately sets a condition - that this sum should be organized from above, that is NOT infinite. Why cannot the Fourier sum be infinite in number of terms, even if it is limited in frequency by an upper limit? Because the Fourier transform consists of a SMALL OF CRATH HARMONICS and if you cut it off at the top, you cannot use it to represent anything but a PILE of an EXTREMELY PERIODIC function on that very PILE. Do you understand? You cannot.
The problem with your (and many others') understanding is that radio engineers are too at ease with the initial conditions of this whole Fourier pile, and allow themselves to jump from one method to another, without caring about preserving the logic.
I will repeat what LProgrammer told you, what decent modern mathematicians know and what Lagrange et al. said in their time when they opposed to wide application of the Fourier method, and at the same time I will show you your delusions, which consist in PROSECUTING through important conditions. This is nothing new, it is described in Fink's books, in the works of Academician Ageyev, in popular form it was even somewhere on "Computer":
Kotelnikov's theorem is not a theorem because it DOES NOT define the concept of "spectrum". What is a "spectrum" in that theorem? It is in that theorem nothing more than FURIER DIVISION. So the theorem is not a theorem at all, it is just a tautology. I am not the one calling it a tautology. It is stated even in old books of academician Harkiewicz.
You are DEEPly slipping into delusions like "Regarding sinusoids - any curve can be represented as a sum of sinusoids." Yes, it is true, only you are not stipulating that "as an EVERY sum of sinusoids". And Kotelnikov's theorem immediately sets a condition - that this sum should be organized from above, that is NOT infinite. Why cannot the Fourier sum be infinite in number of terms, even if it is limited in frequency by an upper limit? Because the Fourier transform consists of a SMALL OF CRATH HARMONICS and if you cut it off at the top, you cannot use it to represent anything but a PILE of an EXTREMELY PERIODIC function on that very PILE. Do you understand? You cannot.
The problem with your (and many others') understanding is that radio engineers are too at ease with the initial conditions of this whole Fourier pile, and allow themselves to jump from one method to another, without caring about preserving logic.
Who are you talking to here?
Go sleep it off, take a bath and drink a cup of coffee and yadda.
just so you know, radio engineers are the sanest people in the world.
..... Yes, because the Fourier transform consists of the sum of CRATH harmonics and if you make it limited on the upper side (break the series), you cannot represent anything except a Piece of an EXCLUSIVE PERIODIC function on this very piece. Do you understand? You cannot.
The problem with your (and many others') understanding is that radio engineers are too at ease with the initial conditions of this whole Fourier pile, and allow themselves to jump from one method to another, without caring about preserving the logic.
Yes I agree, but I hope you won't deny that if there is a periodic function in this area it will show up in the spectrum. How to use this information is another question. The important thing is that we've detected it. Russian radars that protect our country are built the same way, if something moves it means that it has Doppler effect so it can be seen in spectrum (the main thing is to fulfil all conditions of processing and Kotelnikov's theorem is in the first place because if it is not fulfilled everything will collapse).
You've got a wrong idea about radio engineers. Look around you, computer, telephone, mobile phone, television, radio, tape recorder etc. They made it with their hands. They worked very hard to make the mathematicians' formulas work. And they are not as illiterate as they are made out to be. They know and know how, most importantly they know how to put their knowledge into practice.
Yeah, about the same thing, just not in relation to Fourier, I was talking about a few months ago too. We really want to believe that the price will go the way Fourier tells it to go :)
Sorry for the interruption....
I apologize that I have entered, probably not timely, but I could not keep from it as I have the reflections on the given question on which now I actually work, besides the raised theme so-to-say "is close in spirit". it is in general a preamble...
and actually "ambula"... the cochlea in a human ear (and other mammals) "as though" is capable to carry out function of the frequency filter, and in fact if to generalize "it seems that" represents usual spectrum analyzer. further I think few will find objections to that practically any person (if it not deaf certainly) is capable even in the conditions of raised noise (on production for example) to allocate even very difficult BUT PERMANENT signal, (voice of the partner for example) which level is obviously LESS than level of this industrial noise. it is first ...
second, if you run a quote stream through an amplifier in "accelerated mode" you can clearly hear intermittent components that are even more numerous than the noise.
Summarizing all abovementioned it seems reasonable to use Fourier transform for splitting the kotir into a spectrum, with the following giving of this spectrum to the NS input for decision making (NS) - down or up.
PS. i.e. to plagiarize an idea from nature to one's own ends. :)
Yeah, about the same thing, just not in connection with Fourier, I also said a few months ago. We really want to believe that the price will go the way Fourier tells it to :)
Sorry for the intrusion. I read almost the whole thread and couldn't understand the essence of Fourier argument. The subject is the description of market conditions by a small number of parameters influencing the future price movement. What does Fourier have to do with it? I agree that the price movement can be decomposed into sines and cosines: m+An*cos(wn*t)+Bn*sin(wn*t). So? The spectrum (An+j*Bn) will be our description of the market state? The idea is interesting. But in the discrete Fourier transform the number of sines and cosines equals the number of prices taken. What then is the advantage of using the output parameters of the DFT (An and Bn) to describe the market? The number of variables is not reduced. So we should take the largest amplitudes sqrt(An^2+Bn^2). They with their frequencies become the market description? Am I going in the right direction? Using these parameters (An, Bn, wn) we will predict the future by extrapolating the corresponding sines and cosines into the future? Have done such a thing. There is a great misconception in this approach. The Fourier transform is nothing more than fitting a trigonometric series to the original price curve. It makes as much sense as fitting polynomials and other functions to a price curve. You can twist it and take Bessel's functions, sinc, Si and so on. All these adjustments will reach their goal of exact reproduction of the price. But who told us that there are trigonometric functions or polynomials or Bessel functions hidden in the price movement. They are only approximating functions. In order to extrapolate sines and cosines, one must first prove that the price movement is described by ordinary differential equations as an oscillatory circuit. I find it hard to see the benefits of the Fourier transform to describe the market. Although I won't mind if someone decides to change my mind. Who has other ideas?
Exactly. LProgrammer, in referring to Prival to express THIS WAY - used the naughty swear word "b***h". He did this clearly not because he's a swearer and a loudmouth, but to get the attention of Prival and others who are misguided in this area of mathematics.
That is: yes, you can interpolate the tick function of time (process) in some way - in a given interval - but THIS WILL NOT GIVE ANYTHING - as it will not give us EXTRAPOLATION of the process - that is, extension of the function beyond the given interval, predicting it into the future. An exact EXTRAPOLATION can be obtained ONLY by building a proper process model. If you don't have one, you can play with different interpolations until you are blue in the face - each of the interpolations will give DIFFERENT predictions and none of them will be correct. Fourier decomposition into a "spectrum" consisting of a finite sum of multiple harmonics is just ONE of the interpolations. In order to apply the Fourier transform correctly for the purposes of tehanalysis, one has to:
1. to be sure that the function under investigation consists of a sum of sinusoids much larger in amplitude than the noise.
2. be sure that the waveform under investigation has many multiples of harmonics.
Only when these BOTH conditions are met can the Fourier transform sometimes give a correct EXTRAPOLATION. (It can always interpolate correctly - when there are so many harmonic components).