Market theory - page 131

 
Can such charts be shown on Forts - the RTS Index, for example? Because the platforms are different, and then we will be able to understand the nuances.
 
Sergey Petruk:
and why keep "hanging out" the colours all the time?
Now, this is the final version of the colour scheme. if there are any comments, tell us specifically what, needs to be changed.
 
Sergey Petruk:
Can such charts be shown on Forts - the RTS Index, for example? And the sites are different, and then we will figure out the nuances.
It is possible to try, provide data: price at the opening of the session, date, instrument for the last 2 months, TF D1. Or, specify where to get the most reliable data.
 
Yousufkhodja Sultonov:

Yusuf, I asked you a question above. How would you comment on it?
 
Daniil Stolnikov:
So here's a question - MA is known to lag. What can you say about the frequency filter? Does it lag? It shouldn't, but I may be wrong... And isn't MA the same as a filter in its essence?

Allow me to insert my opinion. A filter (as the word itself suggests) is designed to separate useful signal from noise. Here, a filter usually refers to a convolution filter. This operation allows you to manipulate the complex spectrum of the original signal(it multiplies the amplitude and rotates the phase of each frequency).

What do you want to separate from what?

If the useful signal for you is a smoothed signal, you need a low-pass filter. The ideal filter is a two-way filter, you need both past and future to calculate it. A one-way filter inevitably shifts the phase and gives the visual effects of lag, advance, etc., this is unavoidable as you can't know the exact value of the LPF at the present time without knowing the future.

What to do and how to deal with it?

Noise (white noise values, red noise increments) cannot be predicted. But if your signal is not equal-frequency noise and there are steady peaks of certain frequencies, you can filter them out and get (with a certain margin of error) a real-time low pass filter result. See Kalman filter, indicators (there were more, right?).

If you want a simple leading filter, calculate 2*EMA(11)-EMA(22). You can substitute any periods if you wish.

But practice shows that in price data frequency peaks are very weak and unstable and efficiency of such a filter is minimal. So all linear filters are doomed to "lag" or data distortion.

Criticism from professional mathematicians is welcome!

Фильтр Калмана — Википедия
Фильтр Калмана — Википедия
  • ru.wikipedia.org
В большинстве приложений размерность вектора состояния объекта превосходит размерность вектора данных наблюдения. И при этом фильтр Калмана позволяет оценивать полное внутреннее состояние объекта. Фильтр Калмана предназначен для рекурсивного дооценивания вектора состояния априорно известной динамической системы, то есть для расчёта текущего...
 
Daniil Stolnikov:
So far, I'm thinking the same thing.

Yusuf, you are a professor of mathematics, if I am not mistaken? Answer me this question - I have recently started looking into DSP. So here's a question - as you know MA lags. What can you say about frequency filter? Does it lag? It shouldn't, but I may be wrong... And isn't AF the same as a filter in its essence? I cannot answer this question by myself yet.

I, personally, am not a supporter of signal filtering. I use all information by the "as is" principle. The algorithm itself distinguishes between levels with the accuracy of tenths and hundredths of a pip. I do not want to interfere with filtering of the initial signal. What else is needed - without any filtering:


 
Awl Writer:

Allow me to insert my opinion. A filter (as the word itself suggests) is designed to separate useful signal from noise. Here, a filter usually refers to a convolution filter. This operation allows you to manipulate the complex spectrum of the original signal(it multiplies the amplitude and rotates the phase of each frequency).

What do you want to separate from what?

If the useful signal for you is a smoothed signal, you need a low-pass filter. The ideal filter is a two-way filter, you need both past and future to calculate it. A one-way filter inevitably shifts the phase and gives the visual effects of lag, advance, etc., this is unavoidable as you can't know the exact value of the LPF at the present time without knowing the future.

What to do and how to deal with it?

Noise (white noise values, red noise increments) cannot be predicted. But if your signal is not equal-frequency noise and there are steady peaks of certain frequencies, you can filter them out and get (with a certain margin of error) a real-time low pass filter result. See Kalman filter, indicators (there were more, right?).

If you want a simple leading filter, calculate 2*EMA(11)-EMA(22). You can substitute any periods if you wish.

But practice shows that in price data frequency peaks are very weak and unstable and efficiency of such a filter is minimal. So all linear filters are doomed to "lag" or data distortion.

Criticism from professional mathematicians is welcome!

I'm not a mathematician and I'm not interested in white and red noise. From the point of view of practical trading I would like to see a filter that would instantly react (show a change of direction) to only such price movement from which one can make profit. But as this problem is related to prediction and is practically insoluble, I think it is impossible to make profit without using other technical analysis tools, no matter how good the filter is.
 
Yousufkhodja Sultonov:
You can try, provide the data: price at the opening of the session, date, instrument for the last 2 months, TF D1. Or, indicate where to get the most reliable data.
here , openings broker
Files:
 
Yousufkhodja Sultonov:

I, personally, am not in favour of filtering signals. I use all the information according to the "as is" principle. My algorithm itself distinguishes the levels with the accuracy of tenths and hundredths of a pip. I do not want to interfere with filtering of the initial signal. What else is needed - without any filtering:


Yusuf, in the version where you fix daily profits, explain in detail how you calculate the profit. How do you determine the opening and closing price of a trade?
 
khorosh:
Yusuf, in the version where you fix profits daily, explain in more detail how you calculate profits. How do you determine the opening and closing price of a trade?
For example, according to yesterday's trading results we look at the mood of the Lion and place the appropriate order and then close it at the end of the trading day. And in the second case, if the Leo's mood stays the same, we do not close the order, but we give a share in this direction and close all together as soon as the Leo's mood changes.