Is the advisor suitable for real life? - page 4

New comment
 
Mathemat >>:
Точнее сказать так: эксплуатируются не серии прибыльных/убыточных сделок, а серии, в которых профит-фактор сменяется с большего 1 на меньший 1.
Адекватный трейдер при анализе кривой баланса не мыслит единичными сделками (исключения - пипсовщик с SL/TP>>1 и мартингейлер). Он смотрит на график через усредняющее окно - скажем, результат последних 20 сделок. Этот плавающий результат и есть текущий профит-фактор.
Конечно, этой техникой не справиться с сериями сделок ПУПУПУПУПУ... (профит-убыток).


That's what I'm talking about. You need PPPPPPUUUUUUUUUU?
My problem right now is that single losing trades outweigh the profits of a series of profitable trades. The series of profitable trades reaches up to 40 or more trades, but there are very unprofitable ones, which usually occur at reversals or at the end of a trend. I cannot decrease the SL because my Expert Advisor does not cover the volatility. What to do?
 
Mathemat >>:
Точнее сказать так: эксплуатируются не серии прибыльных/убыточных сделок, а серии, в которых профит-фактор сменяется с большего 1 на меньший 1.
Адекватный трейдер при анализе кривой баланса не мыслит единичными сделками (исключения - пипсовщик с SL/TP>>1 и мартингейлер). Он смотрит на график баланса (эквити) через усредняющее окно - скажем, результат последних 20 сделок. Этот плавающий результат и есть по сути текущий профит-фактор.
Конечно, этой техникой не справиться с сериями сделок ПУПУПУПУПУ... (профит-убыток).

So with a series of PUPUPUPUPU.... the balance will stand still?

 
peco писал(а) >>


That's what I'm talking about. Do you need PPPPPPUUUUUUUUUUU?
My problem right now is that single losing trades outweigh the profits of a series of profitable trades. The series of profitable trades reaches up to 40 or more trades, but there are very unprofitable ones, which usually occur at reversals or at the end of a trend. I cannot decrease the SL because my Expert Advisor does not cover the volatility. What should I do?


The strategy like yours is either over sitting or averaging with martingale. It seems to me that it is impossible to profit from it.
For the balance filter to work, profit and loss should be approximately equal. In any case, the period of the short balance average cannot be less than the typical "working cycle" of the strategy.

 
Mathemat >>:
Точнее сказать так: эксплуатируются не серии прибыльных/убыточных сделок, а серии, в которых профит-фактор сменяется с большего 1 на меньший 1.
Адекватный трейдер при анализе кривой баланса не мыслит единичными сделками (исключения - пипсовщик с SL/TP>>1 и мартингейлер). Он смотрит на график баланса (эквити) через усредняющее окно - скажем, результат последних 20 сделок. Этот плавающий результат и есть по сути текущий профит-фактор.
Конечно, этой техникой не справиться с сериями сделок ПУПУПУПУПУ... (профит-убыток).


This is to demonstrate how it works. On a longer period, it's draining.... Is it worth trying lot management here?
Files:
2-3.rar  5 kb
 
Mathemat >>:
Конечно, этой техникой не справиться с сериями сделок ПУПУПУПУПУ... (профит-убыток).

In this case - if the PUPUPUPUP... (profit/loss) gradually increases the balance/equity, the EA can continue to trade.

If the PUPUPUPUPUP... (profit-loss) keep the balance/equity at zero (horizontally) or decrease it, Expert Advisor automatically stops trading and goes to the virtual trading mode, i.e. waiting.
 
Well that's understandable. It's just that peco seems to have a problem with removing unprofitable ones altogether :)
peco, you should at least show me a picture of the balance sheet to get an idea...
 
Dserg >>:
Считаю кривую баланса в пунктах. Строю на этой кривой две машки. Как только быстрая пересекает вниз медленную, уменьшаю лот в 100 раз. Как только пересекает обратно - восстанавливаю лот.
Are any parameters fast and slow pulling the balance north?
Or did you pick them "retrospectively"? Filtering on history is one thing. But picking parameters for your filter for the future is a bit different.
I think that you have a fitting on MM. Although no, reducing the lot by 100 times is ~ a common "fence". Your filter with extra parameters is a regular fit.

Similar, but withmore preference:
Dima_S. >>
In general, the

slope of the linear regression of the balance curve is used, and the adjustment function of the lot size is set depending on the slope...

What do your parameters of the linear regression of the balance curve depend on? On the optics on the history?

 
I have sketched a function to calculate the lot volume for the next order (according to Dserg's suggestion). The only thing is that the lot is increasing/decreasing smoothly.
Variables:
FastPeriodMA - period of fast waving.
SlowPeriodMA - period of slow wave.
Coefficient - multiplier/divider for the next rate.
double GetLot() {
  int i, FastPeriodMA = 7, SlowPeriodMA = 14, Coefficient = 2;

  static double Lot = 0;
  static double PrevBalance = 0;
  static double BalanceOld[0];
  static double BalanceNew[0];
  if (NormalizeDouble(PrevBalance - AccountBalance(), 2) != 0) {
    ArrayResize(BalanceNew, ArraySize(BalanceOld) + 1);
    for (i = 0; i <= ArraySize(BalanceOld) - 1; i++)
      BalanceNew[i] = BalanceOld[i];
    BalanceNew[ArraySize(BalanceOld)] = AccountBalance();
    ArrayResize(BalanceOld, ArraySize(BalanceOld) + 1);
    ArrayCopy(BalanceOld, BalanceNew);
    PrevBalance = AccountBalance();

    if (ArraySize(BalanceNew) > SlowPeriodMA) {
      double FastMA = 0, SlowMA = 0;
      for (i = ArraySize(BalanceNew) - FastPeriodMA; i <= ArraySize(BalanceNew) - 1; i++)
        FastMA += BalanceNew[i];
      FastMA /= FastPeriodMA;
      for (i = ArraySize(BalanceNew) - SlowPeriodMA; i <= ArraySize(BalanceNew) - 1; i++)
        SlowMA += BalanceNew[i];
      SlowMA /= SlowPeriodMA;
      if (FastMA > SlowMA) Lot *= 2;
      else Lot /= 2;
    }
  }
  if (Lot < MarketInfo(Symbol(), MODE_MINLOT)) Lot = MarketInfo(Symbol(), MODE_MINLOT);
  else if (Lot > MarketInfo(Symbol(), MODE_MAXLOT)) Lot = MarketInfo(Symbol(), MODE_MAXLOT);
  return (Lot);
}
Dear experts, share your ideas about MM in terms of speed of increasing/decreasing the lot volume.
Can we mathematically derive a universal formula?
For example, the rate of lot volume decrease in case of a loss should be so much faster than the increase of the lot volume in case of profit.

Remind a binary person, which formula/principle is used to calculate the linear regression (the Dima_S. method) and the subsequent calculation of slope angle of aproximating line (trend). =)
 
coaster писал(а) >>
Do any fast and slow parameters pull the balance northwards?
Or did you "backwards" pick them up? Filtering on history is one thing. But picking parameters for your filter for the future is a bit different.
I think that you have a fitting on MM. Although no, reducing the lot by 100 times is ~ a common "fence". Your filter with extra parameters is a regular fit.

Similar, but withmore preference:

What do your balance curve linear regression parameters depend on? On the optimum on the history?


This is business as usual:
backtesting on history, forward out of the optimization zone.
Otherwise it's not fair, why cheat yourself? :-)
 
Mathemat писал(а) >>
Well that's understandable. It's just that peco seems to have a problem with removing unprofitable ones altogether :)
peco, you should at least show me a picture of the balance sheet to get an idea...


Exactly.
The period is not taken from the ceiling, but from a typical trading cycle.
1234567891011...52
New comment
Reason: