Random Flow Theory and FOREX - page 32
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
After all, the sampling rate is related to the sampling period by the formula Fdisk=1/delta_t. Delta_t is nothing more than the data period (in maths terms "tick lag"). Ask him if tick lags is a random variable (as long as the type of distribution law is not important). If the mathematician says YES, then answer the sampling rate will also be a random variable?
I don't know DSP. Just trying to maintain some notions so that I can orient myself in time if anything happens :)
Tick lags is a random variable, but that in itself doesn't mean the sampling rate is random. It may well be constant, it's just that when the result is the same as the previous tick is not given. In general I think that in a literal sense there is no sampling rate for the market, but we can try to think of the term as an effective concept reflecting the speed of information processing by market makers. From this perspective, imho, we can consider the sampling rate during the working week as constant. The situation with weekends is unclear - I do not know in what mode market makers work at that time, but in any case the uncertainty in determining the "true price" increases at the expense of falling statistics. That is, for one reason or another, the measurement error at market opening on Monday is dramatically increased. That is, most gaps can be attributed to measurement error - I see confirmation of this in the fact that the market in this majority of cases first closes the gap and only then decides where to go next. As for the news, we can assume that at those moments this conventional frequency of sampling is not enough, i.e. we have again the growth of uncertainty in the price determination (this time for another reason) and, as a consequence, the subsequent bumpiness.
That's my imho on the matter.
2 Mathemat: After reviewing Feigenbaum I once again thought about the fact that pseudo-random sequence looks random by all statistical tests being completely predictable. By the way, your future synthetics will also be predictable for this reason :)
Лаги тиков случайная величина, но само по себе это не означает случайности частоты дискретизации. Она вполне может быть постоянной, просто когда результат измерения совпадает с предыдущим тик не даётся. Вообще я думаю, что в буквальном смысле частоты дискретизации для рынка не существует…
Couldn't resist! That's right, what I wrote about is not a property of the original signal, but is literally assigned by the observer to digitise based on the required quality of the original signal representation. But it is the "X" axis, and we can also remember the problem of quantisation of signal, i.e. the "Y" axis, but it is not a property of the original signal either. It is all determined by the combined characteristics of the experiment and simply by the hardware capabilities.
It corresponds to the bit depth of the ADC and for the market it corresponds to a resolution of 1 point. The final resolution gives additional noise.
This corresponds to the bit depth of the ADC and for the market it is 1. The final bit depth gives additional noise.
In our case, the ADC is completely defined and there is no way we can change it, in the sense of making it more accurate. Rougher - no problem. By the way,Candid, do you remember why we need this ADC?
In our case, the ADC is completely defined and there is no way we can change it, in the sense of making it more accurate. Rougher - no problem. By the way,Candid, do you remember why we need this ADC?
Yes, they have no other writers for us :). On a more serious note, it's supposed to be the only accurately calculated source of noise so far. Unless of course my assumption that the DSP has solved this issue is correct :)
To Candid
I was asking globally. :о) Just a bit taken aback by the enormity of Prival's proposal. I don't think that the DSP approach in such a classical sense will help to understand the market structure and emulate it. I'm sure it's a misconception. As for noise, my humble understanding is that this noise does not exist as a class because of the different nature of the source. Yes, there may be "quantization errors" but there is no noise. Ok, let's wait for the author's patient explanation.
to Mathemat
Hurst, Prival? If so, I don't study it too much, but I am definitely going to take it into account when generating synthetics.
And this is much more important than Nyquist frequencies and other nonsense that doesn't work. I strongly recommend you to do this thing, and not only for generating synthetics. Here is a book: "Signal Processing with Fractals: a wavelet-based approach".
http://grasn.narod.ru/002.djvu
I think it may be useful for streams generation as well, but we should remember that Hurst exponent is also a function.
Folks, that's where we are going wrong. What do we care about quantization noise (fractions of a point?) when it is much smaller than the effect produced by DC filters (on the order of a few spreads)?
If we need to generate a synthetic series, we are going the wrong way. It cannot be approached from the position of classical DSP architecture: "Encoder" - "DSP device" - "Decoder". Have fun though :o)
As for noise, my humble understanding is that this noise does not exist as a class because of the different nature of the source. Yes, there may be "quantisation errors", but there is no noise.
As for noise, my humble understanding is that this noise does not exist as a class because of the different nature of the source. Yes, there may be "quantisation errors", but there is no noise.
OK, let me ask it another way. Here's a fractal tree - where's the noise in it?