Log in

ASCtrend Buy Sell expert with artificial intelligence

DOWNLOAD FROM HERE

This is another experiment with the Asctrend Buy Sell expert, kernel optimized by the native MT4 genetic algorythm. The results from build to build in the genetic optimization may vary, build 226 and 409 work both fine.

As there are two kernels you need to optimize them both:

x1 y1: gaussian kernel 1: from -3 to 3 with step size 0.01

x2 y2: gaussian kernel 2: from -3 to 3 with step size 0.01

The risk is central and also need to be optimized: from 1 to 20 

When you look at the results if there are several outliers at the top, neglect them, you need many positive results with the same result on the tester.

Also if you see on the optimization graph mainly negative results you should be aware that the genetic is stuggling to find good solutions.

This expert is designed to cream some more pips over the basic version when the basic version tends to be profitable. If the basic version is not profitable I doubt that the kernels will help you a lot.

Now it was trained from 01.01.2012 to 29.02.02012. 

Out of sample: 01.03.02012 - 25.03.02012

 

image
От Поле за пускане

 

image
От Поле за пускане

 

Comments

  • francisfinley 2765 days ago

    looks interesting. why does it plateau if it self optimises via AI? shouldnt it keep going and going?

     

  • JohnLast 2765 days ago

    No it is not self optimized. Those are out of sample results.

     

  • francisfinley 2765 days ago

    what is the AI function doing? 

  • JohnLast 2765 days ago

    double kernel()
    {
    double w1 = MathExp(-x1*x1-y1*y1);
    double w2 = MathExp(-x2*x2-y2*y2);


    int risk;
    int len = 3+risk;
    double smin=Low[Lowest(NULL,0,MODE_LOW,len,1)];
    double smax=High[Highest(NULL,0,MODE_HIGH,len,1)];

    double p1 = (smax - smin);


    return((w1*p1+w2*p1 )/2);
    }

    This is the kernel function. 

    double w1 = MathExp(-x1*x1-y1*y1);

    http://www.youtube.com/watch?v=ZI67sr0RXzU

    we optimize the parameters x1,x2,y1,y2 here.

    p1 is equal to the difference between 

    High[Highest(NULL,0,MODE_HIGH,len,1)]; and Low[Lowest(NULL,0,MODE_LOW,len,1)]; 

    And we do 

    w1*p1+w2*p1 )/2 

    We multiply this reading by the weight of the kernel, as we have two kernels we devide the result by two.

     

  • isuborsky 2730 days ago

    If I understand correctly this  EA need to optimize or can use set file in folder. I understand this a little

  • JohnLast 2730 days ago

    For the last month there were so little trending action that this system was practically of no use in out of sample.

    It is an expert as any other experts.