Demonstrating the cluster approach to market... - page 18
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
And why the DFT and not the FFT?
Well, that's pretty clear. In FFT, the window (or number of samples) is degree 2. And you may have to play with different lengths.
Why to limit yourself and at the same time to enter the given periodicity where it is not present.
No one can predict the fate of marketmovements. What is left for us to do ?
If we assume the assumption that any imbalance in the base/quote currency prices will be eventually corrected by the market, then this is already a good market strategy.
Trying to predict price movements using various graphs is a futile effort.
This is the strategy of the market. To open every time this imbalance appears and to close every time this imbalance is eliminated by the market.
Trying to predict the movement of prices using all sorts of graphical drawings is a hopeless idea.
+1
Prival писал(а) >>
Берем выборку, желательно с максимальной частотой дискретизации (тики). Учитываем что частота дискретизации не является константой. Выравниваем частоту. Отсеиваем шумы квантования. Строим спектр, желательно не БПФ, а ДПФ. Применяем окно, хеминга, хенинга … и т.д. дальше исследования. Выбираем допустим 3 или 5 или 10 составляющих спектра имеющих максимальную амплитуду, остальные обнуляем. Обратное преобразование Фурье, и получаем кривую, строим её в будущее, анализируем её прогнозные свойства (ветка правила хорошего тона рисунок где 45 градусов) и по кругу пока не найдете приемлемый результат а лучше множество результатов и переключатель между (если в спектре есть ….. то работаем этим адаптивным фильтром, если что-то другое то следующим).
More about all the wrong things. Preferably not just labels, but with justification.
1. There is no need to align the frequency. We're working with ticks, aren't we? Or maybe I'm missing something...
2 Quantization noises do not need to be eliminated either. They will fade away in the process. This is how I do it.
3. Neither FFT nor DFT will do. Or calculate harmonics in some other way using these methods.
4. You can't use just any individual harmonics. Only all up to a period of 4 bars. Anything less is pointless. This will weed out the quantization noise.
5. Using an inverse Fourier transform... Well, anyway, you don't have to do ticks for that. You'll be getting the previous sample all the time. It's a dead end.
Opening whenever this imbalance occurs and closing whenever the market closes this imbalance is the market strategy.
Trying to predict the price movement by various graphical constructions is a futile endeavour.
+1
As long as you like it, I'll repeat two more points that nobody paid attention to.
1. The graph of one particular instrument on which we focus and with all seriousness, carefully
And we conscientiously look for trends, flats, support, resistance, etc.,
represents only the form in which the imbalance of
values of the base currency and the quote currency.
Observing only the shape changes, we will not find any driving force.
-----------------------------------------------------------------------------------------------------------------------------------------
2. At periods of time when we observe on this form
- Flat - there is a real process of alignment of the full-market values of the base currency and the quote currency
This can be clearly seen on the indicator line, reflecting the difference between the full-market values of the base currency and quote currency - line,
At these time intervals the line increases to zero, while the price line of the instrument makes sideways movement.
- Trend - the market starts to increase the imbalance of full-market prices of base currencies and quotes - the line of the indicator
starts to run away from zero.
-----------------------------------------------------------------------------------------------------------------------------------------
I think it is very convincingly shown that the chart of the instrument itself is only a reflection of the full-market
driving forces, the nature and source of which is unknown to us.
I cannot help demonstrating in the picture how Flat and Trend appear:
(the white indicator line shows the difference between the cluster value of the base currency and the quote currency)
.
1. There is no need to equalise the frequency. We're working with ticks, aren't we? Or maybe I don't understand something...
2 Quantization noises do not need to be eliminated either. They will fade away in the process. This is how I do it.
3. Neither FFT nor DFT will do. Or calculate harmonics in some other way using these methods.
4. You can't use just any individual harmonics. Only all up to a period of 4 bars. Anything less is pointless. This will weed out the quantization noise.
5. Using an inverse Fourier transform... Well, anyway, you don't need to do ticks for that. You'll be getting the previous sample all the time. This is a dead end.
1. the arrival time is a constant ? if not, then the frequency is a constant ? if not. Just try, on a model, to reconstruct a simple unnoised sine wave, where the frequency is a random variable. Can you do it ?
2. you can do it without screening, but if I know for sure it's noise, why not.
3. I agree with you there may be different ways, same wavelets for example, but there are some pitfalls too.
4. We obviously count differently here; I count the period in time units and it is in no way related to bars for me.
In addition to the fact that the market itself is a complex function, you also work with bars, which in effect is a non-linear (irreversible transformation) of ticks.
You think they made the terminal especially for Semen Semenych for nothing? (I may be wrong but I bought it for what I sold it for).
And we continue to "wait" for the indicator value to approach zero on any instrument - only then do we close this position....
.....
Here we go.....
And we continue to "wait" when the indicator value is close to zero for any instrument - only then we will close this position....
1. is the arrival time a constant ? if not, is the frequency a constant ? if not. Just try on a model, reconstruct a simple unnoised sine wave, where the frequency is sampled as a random variable. Can you do it ?
2. you can do it without screening, but if I know for sure it's noise, why not.
3. I agree with you there may be different ways, same wavelets for example, but there are some pitfalls too.
4. We obviously count differently here; I count the period in time units and it is in no way related to bars for me.
In addition to the fact that the market itself is a complex function, you also work with bars, which in effect is a non-linear (irreversible transformation) of ticks.
You think they made the terminal especially for Semen Semenych for nothing? If you do not know exactly what you are getting into (I may be wrong, but you get what you get for what you sell).
1. What difference does it make when the tick came! We work with ticks! For us there is some event and it doesn't matter when it happened. We may speak about the number of ticks per time unit or the density of ticks...
2) Noises will be filtered out by themselves. We will still do spectral analysis. It is not difficult to do. It is a side effect of spectral analysis.
3. Yes, there are problems everywhere. The filtering method I came up with has its limit of use too.
4,5. Isn't one tick a bar? It is an equal volume bar in one tick. That's how I see it.
6. Everything I do brings me closer to my own terminal. Thinking about it for the fourth year.
I would like to address this forum's guru.
I have written a multi-currency EA, and the conditions for entering a trade are simple.
But here we have a mishap:
1. The indicator itself is displayed in the chart windows.
2. When compiling the EA, there are no errors with the code.
3. But when you run it in the tester, the deal is not opened and the journal says:
2009.08.14 13:20:18 Multicurrency EA on Cluster Indicator
2009.08.14 13:20:18 2009.01.27 00:00 CL1i_2 EURUSD,H4: loaded successfully
2009.08.14 13:20:18 2009.01.27 00:00 CL1i_2 EURUSD,H4: Init() ended..............
2009.08.14 13:20:18 2009.01.27 00:00 CL1i_2 EURUSD,H4: zero divide
2009.08.14 13:20:18 2009.08.11 23:59 CL1i_2 EURUSD,H4: removed
I think there is a problem with the indicator (I translated it by the translator - zero divide somewhere, if I translated it correctly).
If I try to use other indicators or groups of indicators, the Expert Advisor in the Strategy Tester works, I blame the indicator (maybe I should), I tried to prescribe indicators in the code of the Expert Advisor and 2, 3 and 4 versions.
The answer of the tester in the log is the same.
What is the problem?
I wrote a letter to the author, it took 5 days.
I would like to address this forum's guru.
I have written a multi-currency EA, and the conditions for entering the trade are simple.
But here we have a mishap:
1. The indicator itself is displayed in the chart windows.
2. When compiling the EA, there are no errors with the code.
3. But when you run it in the tester, the deal is not opened and the journal says:
2009.08.14 13:20:18 Multicurrency Expert Advisor on Cluster Indicator
2009.08.14 13:20:18 2009.01.27 00:00 CL1i_2 EURUSD,H4: loaded successfully
2009.08.14 13:20:18 2009.01.27 00:00 CL1i_2 EURUSD,H4: Init() ended..............
2009.08.14 13:20:18 2009.01.27 00:00 CL1i_2 EURUSD,H4: zero divide
2009.08.14 13:20:18 2009.08.11 23:59 CL1i_2 EURUSD,H4: removed
I think the problem is in the indicator (translated by the translator - somewhere in the division of zero, if correctly translated).
If I try to use other indicators or groups of indicators, the Expert Advisor in the Strategy Tester works, I blame the indicator (maybe I should), I tried to prescribe indicators in the code of the Expert Advisor and 2, 3 and 4 versions.
The answer of the tester in the log is the same.
What is the problem?
I have written a letter to the author, it took 5 days.
If anything, I can look at it. It's a division of zero. Often happens in multicurrency, especially in tester mode.