Help write an expert - page 7

New comment
 
Figar0 писал(а) >>
The strategy has a small point, a breakthrough and its confirmation by breaking through the extremum, but 15M timeframe is too small for such a strategy. 1H or 4H IMHO. And of course all these EAs should be improved for the real account.

I hope to do at least a rough test in a couple of days, I also think that M15 is not enough, but I started with small and go ahead...

 
Fduch писал(а) >>

I liked the design of the T.Z. =)

Alesander, your expert is behaving perfectly, no more hooliganism)) Thank you very much for your interest in the TOR;) And seriously, really, thank you very much! I am very glad that my question has attracted many good people to this branch) I will send my Expert Advisor for optimization, and time will show)

 
Fduch писал(а) >>

I advise to leave at least a month for OOS when optimising

I still do not know what OOS is and also, you will laugh, I cannot insert a separating comma to turn a lot from 1 to 0.1. As you can see, it worked here, but not in the Expert Advisor properties. What am I doing wrong, please advise?

 
Vadimus >> :

I wouldn't bother so many people just to give them a textbook example for programming)

What's your reason? Well, you tested it manually and at some point you got a profit, is that a reason? If you had tested it for at least 1 year. If such simple strategies were profitable, all programmers would have become millionaires. I searched through similar strategies when I was mastering MQL and became sure of their futility.


 
khorosh писал(а) >>

What grounds do you have? Well, you have tested it manually and on a certain area and got a profit, is that a reason? If you had tested it for at least 1 year. If such simple strategies were profitable, all programmers would have become millionaires. I searched through similar strategies when I was mastering MQL, and I got convinced of their futility.

And what else could be the reason? Manual testing for a year is possible, but not very useful, because by changing just one parameter you can get a completely different result, which means that you will have to do it manually for a year, and then again .... It takes 10 years of hard work to get an objective picture. That's why people invented MTS, testers, optimisation, etc. It seems to me that I did the right thing. Or do you disagree with me?

 
khorosh писал(а) >>

What grounds do you have? Well, you have tested it manually and on a certain area and got a profit, is that a reason? If you had tested it for at least 1 year. If such simple strategies were profitable, all programmers would have become millionaires. I dug such strategies when mastering MQL, and I got convinced of their futility.

As for futility I would be glad if you tell me where I'm wrong... One of the results of optimization for the period of three months:

profit 2047 pips, profitability 1.98, expectation 22.02, drawdown $522, % drawdown 4.68. Is it absolutely bad. Let all in life will not be so beautiful, but all the same, that the same will be for profit. Or am I wrong?

 
Vadimus >> :

I still haven't found out what OOS is and also, you'll laugh, I can't insert a separator comma to turn a lot from 1 to 0.1. As you can see, it worked here, but not in the Expert Advisor properties. What am I doing wrong, please advise?

This testing is outside the scope of time optimization (Out of Sample, if you translate it from English).

In general, I want to point out that the algorithm of crossover of muwings in both EAs is incorrect. It describes not only crossing but also touching of muwings with their consequent divergence without crossover. There must be false entries - i.e. not provided by the strategy.

The crossover algorithm needs to be modified somewhat. On the fingers: (as yet no time to write in full, if no one will do - I'll write) To determine the crossing from bottom to top of the long moving average short: Must take the difference of the moving average short minus long and compare with a value equal to half the size of a point (0.5 * Point). If more - return true, if less - false. 

bool IsFastUpper(double fm, double sm)

{

    double d = fm- sm;

    if( d>0.5*Point) return(true);

    return(false);

}

Пересечение будет на том баре на котором невыполнение условия сменится на выполнение. 


bool IsCrossedDown2Up(double fm[],double sm[], int i)

{

return(  !( IsFastUpper( fm[ i+1], sm[ i+1]))&& IsFastUpper( fm[ i], sm[ i]) );

}

For other intersections in the same way.


Good luck.

ZS. Regarding the comma - see if you need to put a dot ? it depends on the national standards.

PPS Missed one parenthesis - corrected.....

 
Vadimus >> :

I still haven't found out what OOS is and also, you'll laugh, I can't insert a separator comma to turn a lot from 1 to 0.1. As you can see, it worked here, but not in the Expert Advisor properties. What am I doing wrong, please advise?

Yes, you have. I fix it in each expert!

And OOS: Out Of Sample - outside of optimization interval (on data, which TC has not seen). To estimate what the TS is really capable of, a certain period of time is left without optimization. After optimization the system is trued at this interval. This is how the real is simulated.

Files:
emaccross_2.mq4  4 kb
 
Fduch писал(а) >>

Yes, indeed. In every EA I have this corrected!

And OOS: Out Of Sample - outside the optimization interval (on data that TC has not seen). In order to estimate what the TS is really capable of, a certain period of time is deliberately left out of the optimization period. After optimization the system is trued at this interval. Thus, the real one is simulated.

I'm grateful to you again)

 
VladislavVG >> :

In general I want to note - the algorithm of crossing of muwings in both EAs is made incorrectly.

The fact that there is a "greater than or equal to" sign does not mean that there will be false entries. The matter is that the confirmation signal - penetration of High/Low means that the price keeps moving in the same direction and therefore the moving averages are crossing rather than touching.

123456789101112
New comment
Reason: