From theory to practice - page 430

 
Evgeniy Chumakov:

Either my calculation is wrong, although Alexander seemed to have looked at it.

2. or wrong algorithm for order closing (Alexander was too lazy to explain, so I did as I did).

3. Maybe this strategy only works on ticks and a certain reading interval.


In a month Alexander will tell me how it was on ticks.


p.s. Perhaps, in the formula 3*(SUM(ABS(return))/sqrt(240)) this period 240 should be calculated dynamically, as I can't suggest.

The period should not be touched as it does not really include only time. As I already showed Alexander, for a DC with 4-digit quoting ABS(return) = 10, and this formula gives (N - number of ticks per period):

3*(SUM(ABS(return))/sqrt(240)) = 3*(10*N)/sqrt(240)) = 30*N/sqrt(240) and is rigidly bound to the number of ticks in 4 minutes. If we take 40 minutes instead of 4 minutes, N will grow 10 times and sqrt(2400) will grow not 10 times but 3.16 times, we will obtain quite different values.

 
Renat Akhtyamov:
Look at the last trade, what's wrong with it?


Either closed with the current drawdown or the price crossed the muving and the trade closed on a close condition. I clicked a stop in the tester, took a screenshot and exited.


Where can I read about the law of the root of t that Alexander keeps talking about?

 
Evgeniy Chumakov:


Either closed with the current drawdown or the price crossed the muving and the trade closed on a close condition. I clicked a stop in the tester, took a screenshot and exited.


Where can I read about the law of the root of t, which Alexander kept talking about?

Exactly what I wrote about.

The program is blind.

Read above about what K2 undertook afterwards.

 
Evgeniy Chumakov:

Where can I read about the law of the root of t that Alexander keeps talking about?

Here for example:https://ru.wikipedia.org/wiki/Винеровский_процесс

D[Wt]=t, for RMS will be the root of t

 
bas:

Here for example:https://ru.wikipedia.org/wiki/Винеровский_процесс

D[Wt]=t, for RMS will be the root of t


Thank you!

 
Evgeniy Chumakov:

Where can I read about the law of the root of t that Alexander keeps talking about?

For forex and Alexander's encounter, see https://www.mql5.com/ru/forum/221552/page123#comment_6306015. Unlike the Wiener process model, ergodicity is not required. Exactly on this distinction of local properties of a random process from integral ones https://www.mql5.com/ru/forum/221552/page73#comment_6203173 is, as far as I understand it, the basis of Alexander's approach. A process with no memory (no aftereffects, Markovian) on a large time interval may well have a local memory on small intervals.

The square root law (SQL) is observed in very many real phenomena https://www.mql5.com/ru/forum/220237/page9#comment_6129706, not only in Brownian motion. In https://www.mql5.com/ru/forum/193378/page2#comment_5111351 there are explanations about the difference between forex and Wiener models, there is also a hypothesis about the reasons of the difference, which fit perfectly with Alexander's method of looking for the largest deviations. The effectiveness of the EQC for assessing market activity is shown in https://www.mql5.com/ru/forum/221552/page19#comment_6168925.

 
Vladimir:

Applied to forex and Alexander's encounter, see https://www.mql5.com/ru/forum/221552/page123#comment_6306015. Unlike the Wiener process model, ergodicity is not required. Just on this distinction of local properties of a random process from integral ones https://www.mql5.com/ru/forum/221552/page73#comment_6203173, as far as I understand it, is the basis of Alexander's approach. A process with no memory (no aftereffects, Markovian) on a large interval can well have a local memory on small intervals.

The square root law (SQL) is observed in a great many real-world phenomena https://www.mql5.com/ru/forum/220237/page9#comment_6129706, not just Brownian motion. In https://www.mql5.com/ru/forum/193378/page2#comment_5111351 there is an explanation of the difference between forex and Wiener models, there is also a hypothesis about the reasons for this difference, which fits well with Alexander's methodology of looking for the largest deviations. The effectiveness of the EQC for assessing market activity is shown in https://www.mql5.com/ru/forum/221552/page19#comment_6168925.

ZOK may well hold for a non-stationary process as well (also non-locally, on average). Surely it holds if the distribution of increments exists in an asymptotic sense.

Also, a nonstationary process with Gaussian independent increments may well give a non-Gaussian sampling distribution.

If there is non-stationarity and memory it is not clear what to do with it - we know that the process remembers something but we do not know what exactly. For example, if the word "if" occurs in a sentence, it is very likely that the word "that" will also occur in the sentence. Such an order cannot be explained by Markovian processes but can be explained by stochastic grammars.

 
I'm running the tester with different settings, I can't get positive results, and the whole problem is that the middle is sliding behind the price. Yes, we are back to the middle, but the middle is already down from the entry point.
 
Evgeniy Chumakov:
I am running in the tester with different settings, I cannot get positive results, and the whole problem is that the middle is sliding behind the price. Yes, we are back to the middle, but the middle is already minus from the entry point.

Let's try it together

Send me an indyuq on my post

 
Renat Akhtyamov:

Let's try it together.

Throw me an indium in my post


The formula has already been written here many times.


double SummaReturn = 0;   // Сумма приращений в окне 240 минут
double SummaReturnAbs = 0;

for(int i = 0; i < 240; i++){
SummaReturn = SummaReturn + ( iOpen(NULL,PERIOD_M1,i) - iOpen(NULL,PERIOD_M1,i + 1) );
SummaReturnAbs = SummaReturnAbs + ( MathAbs( iOpen(NULL,PERIOD_M1,i) - iOpen(NULL,PERIOD_M1,i + 1) ) );
}


double Interval_Upper = (3 * (SummaReturnAbs/MathSqrt(240))); // верхний доверительный интервал
double Interval_Lower = -(3 * (SummaReturnAbs/MathSqrt(240)));  // нижний доверительный интервал


Here is the code. This is what is on the second chart from Alexander. I will not publish the third chart without Alexander's permission, because he hasn't written anything about it here.