What is the optimum depth of history for identifying a useful signal? - page 6
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
1- Prediction based on it, extrapolation I meant, otherwise why even lay it out...
2-As closer to the body, but more specifically?
1. That's the whole point. Most people don't know why they should unfold. They are sure that when you have to do something with a signal, you have to do the FFT first. What to do next, they don't know and they don't care, the main thing is to do the FFT. You know that you cannot extrapolate the FFT result unless you want to get an exact copy of the same signal in the future. And if you want to, why did you have to make a FFT? A copy can be drawn anyway.
2. The specifics depend on the specific purpose and conditions. In general, I'm more comfortable talking about time rather than the number of bars. For example, you are interested in volatility change within a day. Don't you think that we should look exactly 1, 2, 3 days, while 1.5, 2.5 days will give wrong result?
1.You know that you cannot extrapolate the result of a FFT unless you want to get an exact copy of the same signal in the future. And if you do, why would you want to do a FFT? A copy can be drawn anyway.
2. Would 1.5, 2.5 days give the wrong result?
1- Why not, you can, with or without extrapolation, but using a decomposition. Well, it depends on how to apply the DFT, of course in the classical variant it will not work, but we are not talking about the classical variant
2- Why is it wrong and wrong in relation to what, is it the only right one? Daily volatility has some properties, but take a 2p 4p window for a sys
But take a 3p 5p window for a syss trough, and if there is a periodicity there will be peaks and troughs but their amplitude will be maximal in the window with more similar size to this "periodicity".
But it is not equal to a day at any given time.
1- Why not, it is possible, both with and without extrapolation, but using decomposition. Well, it depends on how to apply the FFT, of course in the classical variant it will not work, but we are not talking about the classical variant.
2- Why is it wrong and wrong in relation to what, is it the only right one? Daily volatility has some properties, but take a 2p 4p window for a sys
and, as you say, 3p-5p. If there is a periodicity, peaks and troughs will appear there as well, but their amplitude will be maximal in the window with higher values of this "periodicity".
But it is not equal to twenty-four hours at any given time.
1 What are we talking about? We haven't yet agreed on anything and haven't come to any consensus. In the classical version nothing will work, and there can be many non-classical options. I think there is a way to solve the problems, although I haven't got the result yet. But you have no idea what I'm talking about, do you?
2. It should be obvious that volatility is very different between day and night. 1.5 days is day+2 nights or 2 days+night. The results of averaging in these two variants will be different and both will be wrong in comparison with one day. And periodicity has nothing to do with it at all.
Monsieur understands nothing about masochism!
1. what are we talking about? We haven't agreed on anything yet and we haven't come to any consensus. The classical version won't work, and there could be many non-classical options. I think there is a way to solve the problems, although I haven't got the result yet. But you have no idea what I'm talking about, do you?
2. It should be obvious that volatility is very different between day and night. 1.5 days is day+2 nights or 2 days+night. The results of averaging in these two variants will be different and both will be wrong in comparison with one day. And periodicity has nothing to do with it at all.
1. Maybe I don't, maybe I do, how should I know what you think, I also think that there is a way to solve the problem, but I disagree with the interpretation, Fourier has no problems, the problem is with people wanting to stick with its classic version on processes for which it was not created. Correspondingly, the very notion of problems is subjective and cannot be considered as such for all. As I see it, everything rests on finding the optimum level of practicality/speed of calculation. Probably you consider the decomposition in one plane, but with decomposition of residuals, I look at the process as N-dimensional andBut I look at the process in N-dimension and manipulations with measurements are already resource-intensive, and here we have decomposition, the same thing in my variant can be done (decomposition) with FIR filters, even such simple as mashka, I'm not even talking about cf, but computer just can not hold, and to optimize all this, if possible, you should have scaring level of programming knowledge, or kill two or three years for that. So variants without Fourier may be thrown out into open access))) I don't want to kill another couple of years of my life for optimization, I wonder, what will be after announcement of some grail)), and I have enough money besides forex.
There are even more interesting grails, refuting not contradicting terver...including for pseudo sb
There are also options with the reduction of the forex distribution to normal.
2. If we measure volatility not in points, but volumes in combination with points, so much the more, because the daily average tick density is also adequately floating, giving an advantage in some ways than simple bar volatility, especially in multicurrency, one day I will have to tell everything about it in details, I have not got a chance.
1. Maybe I don't, maybe I do, how do I know what you think, I also think that there is a way to solve the problem, but I disagree with the interpretation, Fourier has no problems, the problem is with people wanting to attach its classical version to processes for which it was not created. Correspondingly, the very notion of this problem is subjective and cannot be considered as such for everybody. As I see it, it all comes down to finding the optimum level of practicality/speed of calculation. Probably you consider the decomposition in one plane, but with decomposition of residuals, I look at the process as N-dimensional andBut I look at the process in N-dimension and manipulations with measurements are already resource-intensive, and here we have decomposition, the same thing in my variant can be done (decomposition) with FIR filters, even such simple as mashka, I'm not even talking about cf, but computer just can not hold, and to optimize all this, if possible, you should have scaring level of programming knowledge, or kill two or three years for that. So variants without Fourier may be thrown in the open access))) I don't want to kill another couple of years of my life for optimization, I wonder, what will be after announcement of some grail)), and I have enough money besides forex.
The arguments about the complexity of the calculations and a couple of years I find untenable and far-fetched.
It is not necessary to re-calculate everything on every tick. If the calculations are that heavy, use large timeframes and make the calculation once a day.
There are many people who spend half their lives doing nonsense on forex, so 3 years spent on the case cannot be considered too expensive a price of success.
N-dimensional processes and decomposition of residuals, on the other hand, scare me. It's too complicated. The right solutions tend to be simple.
Arguments about the complexity of the calculations and a couple of years are untenable and far-fetched.
3 - It is not necessary to recalculate everything on every tick. If you have such heavy calculations, use large timeframes and let the calculation be done once a day.
2-There are many people who spend half their lives doing forex nonsense, so 3 years spent on the case can't be considered too expensive a price of success.
1-As for N-dimensional processes and decomposition of residuals scares me. It's too complicated. The right solutions tend to be simple.
3- What makes you think that calculations on every tick meant, another misconception is that if we talk about ticks, then calculations go on every tick.
2- Well they do not know that they are doing a nonsense, they just had bad luck in the search ... depending on what everyone understands success, if you get a certain amount of dough from time to time, then there are options, if it is dough and so comes from another source and its enough, then 3 years of life - to throw away only for more dough, is it a lot or not, if the level of "enough" does not change?
1- Usually, yes (although the criterion of correctness is also a relative concept, if the criterion of correctness is profit, as it should be, why do simple solution has a higher criterion of correctness than the complex, if the complex gives more profitability, or you have only kritari +1 0 -1 profit loss 0, and their value does not matter, then yes, from 2 correct we choose more simple), so they have and performance "normal", if it exists.
A few hundred. This is in case anyone is thinking of trading by the minute signals on the weekly interval (a bit of a tricky decision, I think).