Absolute courses - page 83

New comment
 
Interesting text found http://forum-profit.ru/showthread.php?t=16955
 

There's been a lull... suggest a topic that will generate a lot of emotion and discussion?

Everyone has tried to implement discrete Fourier transform (by the Cooley-Tukey algorithm of the Fast Fourier Transform) in our indices and Expert Advisors. In doing so, the trend (SMA, if you like) distorts the spectral data. Tip: in order to find a "more correct" spectrum, we should not analyze the time series itself, but its derivative. For harmonic components will be preserved during differentiation while trend components will become constants that can be easily discarded.

P.S. From the analysis of differences between "spectra" from the series proper and its derivative, one can draw conclusions about movements that are not "spectral" (harmonic) in nature. Here, colleagues, is the separation of a trend from a flat - the key element of the Grail.

 
Dr.F.:

There's been a lull... suggest a topic that will generate a lot of emotion and discussion?

Everyone has tried to implement discrete Fourier transform (by the Cooley-Tukey algorithm of the Fast Fourier Transform) in our indices and Expert Advisors. The trend (SMA, if you like) distorts the spectral data. Tip: in order to find a "more correct" spectrum, we should not analyze the time series itself, but its derivative. For harmonic components will be preserved during differentiation while trend components will become constants that can be easily discarded.

P.S. From the analysis of differences between "spectra" from the series proper and its derivative, one can draw conclusions about movements that are not "spectral" (harmonic) in nature. Here, colleagues, is the separation of a trend from a flat - the key element of the Grail.


Taking the derivative is equivalent to multiplying by jw (i.e. frequency) in the frequency domain. This means that all low-frequency components will be suppressed, while high-frequency components, where all noise is also found, will be amplified many times over. There will simply be nothing left to analyse (at least much less useful).
 
alsu:

Taking the derivative is equivalent to multiplying by jw (i.e. by the frequency) in the frequency domain. This means that all the low frequency components will be suppressed, while the high frequency components, where all the noise is also found, will be amplified many times over. There will simply be nothing left to analyse (at least much less useful).

Well then do "normalization", divide the spectrum found by a straight line - linearly increasing frequency. somehow. no? In general, I think that you just did not do it in practice. I am even sure of it.
 

A strange behaviour of the EA has been detected. The close order function either closes the selected order or returns an error code. I see in my alerts: "Order USDCAD not closed, error 0". My Gestapo methods let me down, I give you the code.

int close(string sy)
  {
  while(exist(sy)!=0)
    {
    double ask=NormalizeDouble(MarketInfo(sy,MODE_ASK),MarketInfo(sy,MODE_DIGITS));
    double bid=NormalizeDouble(MarketInfo(sy,MODE_BID),MarketInfo(sy,MODE_DIGITS));
    for(int i=0;i<OrdersTotal();i++)
      {
      OrderSelect(i,SELECT_BY_POS);
      if(OrderSymbol()==sy)
        {
        RefreshRates();
        if(OrderType()==OP_BUY)bool res=OrderClose(OrderTicket(),OrderLots(),bid,20,Yellow);
        if(OrderType()==OP_SELL)res=OrderClose(OrderTicket(),OrderLots(),ask,20,Yellow);
        }
      }
    Sleep(3000);
    if(res)Alert("Ордер ",sy," успешно закрыт по смене сигнала");
    if(!res){int err=GetLastError();Alert("Ордер ",sy," не закрыт, ошибка ",err);}
    }  
  return(0);
  }

This has never happened before. But now the EA is stuck in the close function.

 

Here you are, dear alsu, taking the trouble, so to speak...

The effect you mentioned has not been detected.

P.S. See the same little peep somewhere around the 40th "spectral component"? Here it is true. Saved when differentiated. And that the large modulo low-frequency components on the left are from the evil one. It is not true. For in differentiation it is gone. But at the same time, it distorted and redetermined its high-frequency components. Do I get the point?

 
grell:

A strange behaviour of the EA has been detected. The close order function either closes the selected order or returns an error code. I see in my alerts: "Order USDCAD not closed, error 0". My Gestapo methods let me down, I give you the code.

This has never happened before. But now the EA is stuck in the close function.

You have to do the opposite in the loop, from the last order to the first

for(int i=OrdersTotal-1;i>=0();i--)
 
Dr.F.:

Well, then do "normalisation", divide the spectrum found by a straight line - linearly increasing frequency.

You mean reintegrate backwards? Very clever.

No? In fact, I think you just haven't done it in practice. I'm even sure of it.

You're shining with confidence again, just like at the beginning of this thread, aren't you?
 
alsu:

You mean reintegrate? Very clever.

Shining with confidence again, like at the beginning of the thread, eh?


Why shouldn't I shine? I haven't lost my account yet, there will be a profit :-) That the spectrum will be normal is shown clearly to you.
 
Dima.A.:
you have to do the opposite in a loop, from the last order to the first

Prove it!
1...767778798081828384858687888990...113
New comment
Reason: