A quick and free library for MT4, much to the delight of neuralnetworkers - page 31

New comment
 
marker:

I'm only 20% optimised yet (your EA), and the ANN folder is already 1.1 gig!:)) He's eating up space:))) When it finishes optimizing, I'll get in and I'll dig, but how to find everything among so many files.....

Search: in Explorer, type in a partial file name, for example EURUSD35 and get a list of all files in the directory with this sequence in the file name.
 
VladislavVG:

ZS Yes, and a word of advice - don't rush into the real thing just yet. These EAs are more a demonstration of the capabilities of the approach than a customizable skirmish.


You do not use it on real account? "What is that?)) At least there are some clear stops. If i've tried many different junk, martingale, EA with 10-20 pips take and 200 and 500 stops (over sit), so this EA in my opinion is not the worst version.... something like that, if you would like to know something better, i'd be glad.
 
VladislavVG:
Search - in explorer type in a partial file name, e.g. EURUSD35 and you will get a list of all files in the directory with that sequence in the file name.

Thank you:)
 
marker:

You do not use it on real account? I would say "more adjustable than a skirmish" - what is that?) If i've tried it, i've never seen such humps, i've never seen martin, i've seen martingale, i've seen advisors with 10-20 pips take and with 200 and 500 stops, so this EA is not the worst version.... and i'd be glad if you could recommend something better.

No, I don't use it, at least not head-on. But in principle, you could make something out of it. To do so, you would need to do a bit of "tinkering" with the input system. Or just change the whole approach ))))))))).

The entry system of this EA is: RSI with period 30. It's like trying to trade on this indicator alone. 

void ann_prepare_input () {
    int i;
    double res = 0;
    for(i = 0; i < AnnInputs; i++) {
      res = (iRSI(Symbol(), 0, 30, PRICE_OPEN, i) - 50.0) / 50.0; 
      if (MathAbs(res) > 1) {
         if (res > 0) {
            InputVector[i] = 1.0;            
         } else {
            InputVector[i] = -1.0;            
         }
      } else {
         InputVector[i] = res;            
      }
    }
}

No grid will be able to trade it profitably for a long time. It is not difficult to check it - the algorithm is as follows:

1. Do not conduct optimization on the entire available history, but, for example, assume that it is August 2010. Optimize it BEFORE this time.

2. Run the variant you like AFTER the optimization date to the present day.

It is called a forward test and saves a lot of time, rejecting unstable options, and most importantly, money.

And you will get a "customizable setup" when your trades will be profitable for some time and the amount of deals - not just one or two, of course - will be calculated in such a manner that the system will work outside the optimization period. Then the only thing left is to tweak it from time to time ......

 

Well, I know what a forward is, I must have asked that question on this site a year ago, or maybe two already:)) And about the shallow frecht-I cried:))))))))) As for the forward grid (this is not the usual EA), because the result will still be different, because it is a network, it always shows different things (I do not understand why, like specific weights there are always changing, I do not get it), with normal EAs in this regard is much easier....

 

By the way, I never understood the principle behind it, I couldn't read the code, but something about RSI levels:))

 

"Outside the optimization period the system will for some time and the number of trades - not one or two, of course - work in profit" - almost any advisor can be purchased, so that it would at least for some time in the plus, as long as it would not lose and the drawdown was not strong, that is another issue:))

 
marker:

Well, I know what a forward is, I must have asked that question on this site a year ago, or maybe two already:)) And about the shallow frecht-I cried:))))))))) As for the forward grid (this is not the usual EA), because the result will always be different, because it is a network, it always shows different results (I do not understand why, it seems that the weights there are always changing, I do not understand it), with normal EAs in this regard is much easier....

There, in the function start() {}, there is code that adjusts grid (fine tunes) on forward test..... 

   // Adaptive part
   if (IsOptimization() || IsTesting()) {
      total = OrdersHistoryTotal();
      if (total > 0) {
         OrderSelect(total - 1, SELECT_BY_POS, MODE_HISTORY);   
         if (OrderProfit() < 0) {
            if (OrderType() == OP_SELL) {
               train_output[0] = 1; 
            } else {
               train_output[0] = -1; 
            }
            // Learning
            for (i = 0; i < AnnsNumber; i++) {
                       ann_train (AnnsArray[i], InputVector, train_output);
                      }
         
        }
      }
   }

IMHO - it doesn't allow for adequate evaluation. If you remove it, then everything will be the same as in a normal EA.

 
VladislavVG:

There, in the start() {} function, there is code that adjusts the grid (finishes) on the forward test.....

IMHO - it doesn't allow for adequate evaluation. If you remove it, then everything will be the same as in a normal EA.


I agree that it is inadequate, I agree that the specific weights in the run should be the same as in the optimization, otherwise the second run is "from scratch" - as far as I understand it. In the real world to "finish" will not work, therefore, need to rogue on the warrant with the same specific weights, but need to learn how to save them, I've seen somewhere in the branch - it is possible.
 
marker:

"Outside the optimization period the system will for some time and the number of trades - not one or two, of course - work in profit" - almost any advisor can be purchased, so that it would at least for some time in the plus, as long as it would not lose and the drawdown was not strong, that is another issue:))

Not always: Because of the limited possibilities for history analysis, it is possible to gain, for example, during the end of a trend and when moving into a consolidation zone or trend reversal, a lot of money is lost.
1...242526272829303132333435363738...57
New comment
Reason: