Page 4 of 6 FirstFirst ... 23456 LastLast
Results 31 to 40 of 55
  36 36 Attachment(s)    

Thread: PerceptronOscill_v1 EA

  1. #31

    Default

    Sorry. Correction. With shift=1 on 1.02 version and set file for 2009 it makes 13k profit on a 5k account instead of 500k or more.

  2. #32

    Default

    Funyoo or logan113 could you specify which is the good code? the modification made by logan113 a few post back or the original? I want to make some optimizations but i need to know the good one to not waste a lot of time on the wrong one.

  3. #33

    Default

    i made some modification of funyoo's original code ie insert a ema trend filter and modify a bit the perceptron oscillator filter. but both version are good
    actually it's your choice. do an optimization with both ea's with your broker data and use the one that give best results.

  4. #34

    Default GBP/USD optimization

    here are results for gu on perceptronoscill_v1.02

    2009
    profit : +1800%
    RDD : 27.91%
    Attached Images Attached Images PerceptronOscill_v1 EA-perceptronoscill_v1-ea-gu-m5-v1-02-gif 
    Attached Files Attached Files

  5. #35

    Default

    @logan113 Which broker was that backtest used for?

  6. #36

    Default

    Quote Originally Posted by logan113 View Post
    Hi funyoo, i found an error in the code that limited the perceptronoscill filter part of the code as it was using period of the perceptronoscill sigal part.

    Original coding :

    Code:
    if(period<2)period=2;
          if(period>90)period=90;
    
          kk=NormalizeDouble(d/(period+1),0);
      
          double pco=0.0;
          for(c=period;c>1;c--){
             pco=pco+(((kk*c)-(d/2))*(iClose(NULL,signaltf,shift)-iClose(NULL,signaltf,shift+c)));
          }
          double pcoa=0.0;
          for(c=period;c>1;c--){
             pcoa=pcoa+(((kk*c)-(d/2))*(iClose(NULL,signaltf,shift+1)-iClose(NULL,signaltf,shift+1+c)));
          }
          double pco2=0.0;
          for(c=period;c>1;c--){
             pco2=pco2+(((kk*c)-(d/2))*(iClose(NULL,filtertf,shift)-iClose(NULL,filtertf,shift+c)));
          }
          double pco2a=0.0;
          for(c=period;c>1;c--){
             pco2a=pco2a+(((kk*c)-(d/2))*(iClose(NULL,filtertf,shift+1)-iClose(NULL,filtertf,shift+1+c)));
          }
          time=Time[0];
    updated code :
    Code:
          if(period<2)period=2;
          if(period>90)period=90;
          if(period2<2)period2=2;
          if(period2>90)period2=90;
    
          kk=NormalizeDouble(d/(period+1),0);
          double kk2=NormalizeDouble(d/(period2+1),0);
      
          double pco=0.0;
          for(c=period;c>1;c--){
             pco=pco+(((kk*c)-(d/2))*(iClose(NULL,signaltf,shift)-iClose(NULL,signaltf,shift+c)));
          }
          double pcoa=0.0;
          for(c=period;c>1;c--){
             pcoa=pcoa+(((kk*c)-(d/2))*(iClose(NULL,signaltf,shift+1)-iClose(NULL,signaltf,shift+1+c)));
          }
          double pco2=0.0;
          for(c=period2;c>1;c--){
             pco2=pco2+(((kk2*c)-(d/2))*(iClose(NULL,filtertf,shift)-iClose(NULL,filtertf,shift+c)));
          }
          /*double pco2a=0.0;
          for(c=period2;c>1;c--){
             pco2a=pco2a+(((kk2*c)-(d/2))*(iClose(NULL,filtertf,shift+1)-iClose(NULL,filtertf,shift+1+c)));
          }*/  // not used in the ea
    Maybe that should be interesting to make an other optimization with the non limited filter we could get better results

    Ps: V1.02 was not correcting all the filter. a mistake i made (luckily... )
    logan113 i am asking about this modification of code. witch one is the good one?
    in forward test version 1.02 is opening position that are not found in optimization or backtest. this is due to shift=0 on filter. therefore any optimization is without merit because of missing positions.

  7. #37

    Default

    @logan113 Which broker was that backtest used for?
    Hi Agentrfr

    I used a JadeFx terminal with the data's they are providing on their website.

    logan113 i am asking about this modification of code. witch one is the good one?
    in forward test version 1.02 is opening position that are not found in optimization or backtest. this is due to shift=0 on filter. therefore any optimization is without merit because of missing positions.
    Hi daxc

    I'll explain : This ea use for entry purpose indication from two perceptronoscill indicators one working as a entry signal and one for filtering entries. In v1.0 both filter and signal indi were using same period for calculation. wehter there were in the ea settings "period1" for signal and "period2" for filter inputs. So i modified it to use the "period2" input for the calculation of the filter indicator. BUT when i modified the code i made a mistake missing to modify a part of the code. So in v1.02 filter and signal indicators are not totally independent like i intented to do. As optimizations were better like this than with the all modified EA (ie with filter and signal totaly independant) I decided to keep it like this. I also posted the code of the correct modiffication for information just in case someone would like to test it like this. Hope i have been clear.

    For the shift i suppose that the ea is repainting on the last bar. You should add to the code a feature to calculate indis only on the first tick of a bar.
    I optimized it using control point only and results were close to all ticks so it shoudn't be a problem.

  8. #38

    Default

    Thanks logan113 for info. In forward test 1.02 version with 1.02 set has opened 5 transactions in 8.01 11.01 12.01. In back test on the same days has made only 3 transactions. So now i have 40% more transaction that do not appear on back tests and we have no optimization for them. Out of those five one was a loser one a big winner and 3 small gain (breakeven). In my backtests only 10% are looser but if we have in real life more transactions and more loosers what appear to be a high winning program could actually be a lemon. So please everyone proceed with caution on real account.
    If someone has some forward test with 1.02 please compare with backtest for the same period and tell us if you have different transaction in forward vs backtest.

    Thanks again to everyone involved with development.

  9. #39

    Default

    Hi Funyoo

    as this ea act very well in good volatility especially for EU and give a lot of good entries, i thought reading this thread bluestrader-ea that it could be nice to implement this exit strategy instead of the one already existing in the ea. I had an attempt to make it on my own but my coding skills seem to be a little bit poor to do it. could you please code it for me : delete older exist strategy and use Bluestrader one instead. Thanks in advance.

  10. #40

    Default USD/JPY Optimization

    here are the results for UJ :

    v1.02 optimized 01/2009 to 10/2009

    on 2009 to today
    profit : +8300%
    RDD : 44.88%

    Note : the optimization is not compete because of a non desired reboot
    Attached Images Attached Images PerceptronOscill_v1 EA-perceptronoscill_v1-ea-uj-m5-v1-02-gif 
    Attached Files Attached Files

Similar Threads

  1. PerceptronOscill_v1 EA
    By fxtester in forum Expert advisors live statements
    Replies: 1
    Last Post: 04-14-2010, 16:46

Tags for this Thread

alpari, alpari uk, average, backtest, backtesting, bands, breakeven, broker, coding, eurusd, free, funyoo, high, indicator, indicators, live, martingale, mql, oscillator, profit, profitable, real, signal, signals, slippage, sma, spread, stop, strategy, study, systems, test, thank you, time, trading, trading signals, trailingstop, trend, usd

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •