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
Read here for answers to all your questions.
Thank you, Prival.
I've looked through it. I'm sorry, I probably won't read it.
I don't think I can master so many formulas on my own.
.
I hope, I shall satisfy my curiosity simply in matlab.
I've already found a step-by-step tutorial.
Thank you, Prival.
I've looked through it. I apologize right away, I probably won't read it.
I don't think I can master so many formulas on my own.
.
Hopefully, I will satisfy my curiosity just in MathLab.
I have already found a step-by-step tutorial.
And you're wasting your time not reading it. If you knew the theory, 99% of your questions would be gone. But the gut feeling method may be the answer, but it will take a lot of time + you are not sure that the answers will be correct.
Yep...
It doesn't make much sense to process rows with gaps from Soros, 911 and other disasters. You can use filters or Fourier. The question is, if we take those gaps out, what will be left.
Mua-ha-ha...
I wonder if the nonces are included here as well? ;-)
A row is one and there are many candles in it, and when there are many, it's a statistic.
It's about the distribution of candle heights. More precisely, it's about the fact that there are noticeably more very big candles than there should be, if no one is behind it.>> From the same place.
"As a sample of a random variable, a sample consisting of the logarithms of the relative change in the value of the Russian Trading System Index (RTS Index) over
from 1 September 1995 to 31 December 2002."
I ran the LF filters in the matlab simulation.
The phase of the signal after the filter in general case is floating.
For those interested - the simulation model for Matlab 2006 in the attachment.
.
As a result the answer of Kenny & Goodman (authors & customers of the generator at fx.qrz.ru) came to mind
about the fact that as a result of changing the filter parameters the phase "floats".
Including can flip a = -a.
An important fragment got lost - nothing to say.
They justified this answer by peculiarities of the algorithm.
.
Just in case someone might be interested -
the answer to my question is - with low-pass filters fx.qrz.ru - potentially - everything is fine -
but you have to stupidly pick up an LF that "can" handle the task,
for example, a filter in MathLab, calculated by LNA with a cutoff of 110 - 130 Hz
for some signal will be able to filter 100 Hz without phase shift,
but with 30 orders of magnitude it will, and with 200 orders of magnitude it won't.
.
Since in fx.qrz.ru filter order cannot be set forcibly,
then - in theory - it is necessary to "play" with those parameters that are given
to the user (P1 / D1 / A1 / Ripple).
The Market Model for Digital LF Filters
"Smart fleas on a fat dog"
or like the fleas who know nothing of the dog's life, but have a low-pass filter,
to guess the movements of the host and thereby profit from the host's movements.
-Dog asleep
-Dog eats
-The dog itches (do not enter)
-The dog wags its tail (no fleas to bite)
-Dog breeds (you can change the carrier)
-Dog catches fleas (stages: dangerous pause before snapping teeth, fine-pitch plucking)
=All these tasks are solved by a good low pass filter.
These examples can be summarised in terms of one thing in common - the continuity (smoothness) of the process. Indeed, a dog making love cannot suddenly fall asleep or start a hunting session. If the dog scratches, it cannot be bitten immediately, etc. In short, in the examples given, if a process has started then it is likely to continue, this is really a task that can and should be handled by the LPF. Using digital filters, we are entitled to count on their predictive capabilities in this case. This is based on the smoothness of the initial Time Series (positive correlation coefficient between neighboring samples in the first difference series of the initial BP). The series having such properties may be sufficiently smoothed by a mowing and a forecast may be built several steps ahead by any method, for example by decomposing the mowing at the right end of the series into a Taylor series or you may do without any previous smoothing at all, having applied the RT decomposition to the initial BP, it does not contradict the applicability of the method.
As for BP of price type, the picture here is fundamentally different. Let me repeat once again, these series as a rule are alternating in the series of the first difference (not smooth), all the apparatus of diffusion will not work and does not work. It makes no sense to smooth the initial BP and then forecast it - the inevitable FZ during smoothing pushes the mouwing beyond the horizon of a possible future forecast (it's a consequence of the fundamental impossibility to look into the future).
So, dear Korey, the example you cited is not a "Market Model for digital LF filters", but one of many examples for digital LF filters that have exactly nothing to do with the market.
Hmmm... So the problem is formulated (in particular) as a prediction of the next candle's form based on a series of previous values?
Yes. But not necessarily the candlestick's appearance, often it's enough to predict only the colour.
As for the big candlesticks... Based on what assumption was it made that there shouldn't be a lot of them? You mean the shape of the distribution curve?
There are still zeros along the edges, i.e. candlesticks of + or - a million points or more do not exist. But the curve may be of any kind, may it have one maximum or two or more?
The point is that this kind of distribution (with thick tails) is not "convenient" from the viewpoint of possible risks. In general, it is what it is - far from normal, it does not contradict anything. However, there cannot be two or more humps in the distribution. It is related to non-stationarity of such distributions. They can exist only if there are noticeable unidirectional flows inside a closed system. In this case, the realization of a two-humped distribution would require a constant unidirectional flow of money to maintain it... On this, it is immediately possible to make money, i.e. such a market is essentially inefficient, which is contrary to the basic premise of the market.
The filter parameters will cause the phase to "drift" as a result.
Including a = -a can flip.
An important fragment is lost - nothing to say.
Again, that's what I've already said https://forum.mql4.com/ru/10977/page27
in that program just look at amplitude-frequency response and frequency response
With certain parameter combinations, resonances in harmonics and/or phase reversal and other distortions occur.
that is why you have to control the AFC and FFC
Those who have built oscillating circuits by hand to solder them into some equipment know about the vagaries and surprises of our complex world
As for price-type BPs, a fundamentally different picture emerges. Again, as a rule, these series are alternating in the first-difference series (not smooth), the entire diffusion apparatus will not work and does not work. Smoothing initial BP and then forecasting it does not make sense - the inevitable FD on smoothing pushes the mouwing beyond the horizon of possible future forecast (it's a consequence of the fundamental impossibility to look into the future).
I think you're wrong the initial price BP is continuous and hence the diff. calculus works fine. This is a bad quoting system, just because no ticks are coming to us does not mean that the price is missing (has gaps), plus the digitisation of this BP is accurate to the 4th digit.
The fact of a distorted and inaccurate quoting should be taken into account in the first place, especially when dealing with minutes. Then the forecast will start to work out.
certain parameter combinations result in harmonic resonances and/or phase reversal and other distortions
that's why you should control the AFC and FFI.
Looked at it again.
Apparently, you, from your side, know what you are advising - and even use it yourself.
But for me, as a newcomer to filters, combing AFR /FFFR gives me nothing.
In the program itself, the changes are also of little significance to the eye.
So - thank you :-).
But with sine wave at the input everything is clear.
.
P.S.: the program breaks down when method #1 is active.
I think you are wrong the original price BP is continuous and hence the diff. calculus works fine. This is a bad quoting system, just because no ticks are coming to us does not mean that the price is missing (has gaps), plus the digitization of this BP is accurate to the 4th digit.
But we have to and can work with what we have, and it is not smooth in the first difference of BP. Consequently, it makes no sense to use muves in this filing.
The fact of bent and inaccurate quotes should be considered in the first place, especially when working on minutes. Then the forecast will start to work out.