What is the optimum depth of history for identifying a useful signal? - page 4
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
Wow, what about the FFT and the FFT in it.
A useful signal is one that can bring benefits, earnings.
So why the fuck look for it on a 5000 bar history?
Is that a question for me? I don't know, I didn't search for 5000 bars.
he must have meant.
"I am making a predictive indicator. My forecasting problem is reduced to the least moduli method. There are a number of unknowns. We calculate the values of several functions of these unknowns on some number of bars, take modules of all these values and add them all together. The unknowns need to be selected in such a way that the sum is minimal. Suppose the unknowns are 2500, functions are 10 and bars are 5000. Having solved the problem we will get a forecast for 2500 bars".
Never mind...
How many bars were you looking at?
One does not contradict the other.
In its pure form, the DFA is not good for the market. To be honest, I don't know what it is good for at all. There was even a time when I thought it was flawed. Now I think it's just not finished. If you manage to finish it, you might get some very interesting results.
If you are seriously interested in FFT, you should study it not from books but at a higher level. You have to understand the physics of what's going on.
Which frequencies (periods) are in FFT and which are not and why? Why at even N the spectral counts on the left and on the right of Fd/2 are complex-conjugate? Why at even N the spectral counts 0 and Fd/2 have no imaginary part? What is obtained at odd N and why? Most people, even those who consider themselves experts, do not know the answers to these questions.
If you can answer the above questions correctly, more serious questions should arise that should suggest a way of solving the problems. I am in the process of doing this and so far the prospects look very tempting.
Let's drop odd symmetry and even asymmetry that introduces extra (which is not in the signal) frequencies when trying to decompose the remainder.
Let's move on to meanings.
The smoothness of the change in the forecast, isn't that the point.
The question wasn't for you, but you answered it. As it happens: How far into history do you look to get a signal?
I thought I asked this question in the thread))))
You're plotting curves on the screen too, probably on some data that has depth in bars, no?
ZaPutina:
The smoothness of the forecast change, isn't that the point.
That is definitely not the point. What is a prediction, as far as a normal DFS is concerned? There is and can be no prediction there.
The question was not for you, but you answered it. As it happens: How far back in history do you look for a signal?
It depends on what it's for. Exactly for getting a signal it is bothering me to look at more than a hundred bars. But it is useful to look at several thousands to study the behavior of the chart itself.
1- This is definitely not the case. What is a prediction, as far as a normal DFS is concerned? There is and can be no prediction there.
2- It depends on what for. It is for getting a signal that looking at more than a hundred bars is kind of stressful for me. But to study the behavior of the chart itself it is useful to look at a few thousand.
1- Prediction based on it, extrapolation I meant, otherwise why even lay it out...
2- Closer to the body, but more specifically?