Page 1 of 2 12 LastLast
Results 1 to 10 of 20
  10 10 Attachment(s)    

Thread: N7s Ao 772012

  1. #1
    Administrator funyoo's Avatar
    Join Date
    Sep 2008
    Posts
    7,003

    Default N7s Ao 772012

    Here is a backtesting with the N7S AO 772012 expert advisor.

    EU M5. On one year.

    Total net profit : 31.14%
    RDD : 20.27%
    Attached Images Attached Images N7s Ao 772012-strategytester-n7s_ao_772012-eu-m5-gif 
    Attached Files Attached Files

  2. #2

    Default This one relies on weekly optimization

    I saw original in Russian forum. Also Someone by name George posted optimization tests on TSD, but after a month they were all deleted. TSD isn't geeky enough for this. If you optimize it every week it brings steady profits on Euro. I forward tested it, nice risk/reward and %win. But after George was gone from TSD his older .set stopped working. It's a nero net, one needs super CPU computer and a day of patience on weekend to tweak it on last 6 months of quality data.
    **
    and u need lots of tweaking experience and very good data collected from your broker in real time
    ***
    It's a high maintenance EA
    ***
    If you don't know how to tweak MT4 and maintain quality data, move on, this is not 4u
    ********************

    Funyo, you are missing Include folder:
    #include <WinUser32.mqh>
    #include <stderror.mqh>
    #include <stdlib.mqh>
    ************************
    Also this system has optimization .set templates. To optimise variables in cycles, cause EA is so involved.
    Again it is very high maintenance. This is for geeks.
    *************
    Here are all the files. Entire folder system_V2 goes into Experts>Include.
    5-16-2009eu.set was optimized for IBFX no hedge.
    N7s optimisation .set.zip contain optimization .set templates to run optimization in several cycles.
    **********
    System was tested on other pairs, but works best on euro.
    may not be suitable for live trading as trade error checks were nod coded yet. It's a beta version
    Attached Files Attached Files
    Last edited by Metatrader7; 07-26-2009 at 21:59.

  3. #3
    Junior Member
    Join Date
    Mar 2009
    Posts
    6

    Default

    can post the forward test results?

  4. #4

    Default

    Quote Originally Posted by doshur View Post
    can post the forward test results?
    it makes 2-3 trades per week 3 on average and pulled around $500 on 0.1 fixed lot with fresh optimization. Then it started losing more and more the following weeks. So it depends on weekly optimizations as I said and keeps doing well if optimized for the next week or two, sometimes 3. So I shut it down cause I don't have time for that. May do it later, but you have to do it for few weeks to become confident and test your optimization skill. It's very educational. You become a pro optimizer after you do that. Another thing that puzzled me, After I forward tested it and it pulled $500+ in a week, I back tested it on the same period data and back test didn't show anything spectacular. Visual back test showed results somewhat close to forward but not as good. So it was a surprise to see something is better on forward then on back test.

    Russians forward testing it (latest version) already for at least 6 months with positive results on euro. It has trail stop and classy MM with good risk/reward and %wins. No martingale, none of these scalping bs or gambling junk.

    Real Trader EA. I post it as an example that something like that does exist.
    Here is the link to the source, it's all in Russian:
    MultiNeyro - MQL4 Code Base

    Here is English
    http://codebase.mql4.com/5093

    ****************
    Seems Russians don't care that much about others understanding them, but there is a big history of chatter with all tweaks for this system in Russian if you can read.
    ************
    It's like in Architecture - best sky scrapers are in Dubai, not US, although they start building them first. It seems best EA coding is in Russia and if you read Russian, you'll find entirely different level of thought.
    ***
    English speaking sites are dominated by junk, marketing and confused gamlbers, very inferior culture of thought but proactive. People are trained not to think for themselves and follow marketing bs. Whatever works with least effort and most of the time it doesn't, but have just enough fake appearance to sell it for $$. So you go figure.
    Last edited by Metatrader7; 07-28-2009 at 11:00.

  5. #5
    Junior Member
    Join Date
    Mar 2009
    Posts
    6

    Default

    Hi,

    I myself had created a EA with this perceptron filter and was working well also.

    I agree it is time consuming and would like to find out more and learn more about this perceptron thing.

    I share how I code my perceptron here.

    //+------------------------------------------------------------------+
    //| Perceptron |
    //+------------------------------------------------------------------+
    double Perceptron()
    {
    double w1 = (x1 - 100) * PP(0, 7);
    double w2 = (x2 - 100) * PP(7, 14);
    double w3 = (x3 - 100) * PP(14, 21);
    double w4 = (x4 - 100) * PP(21, 28);

    return(w1 + w2 + w3 + w4);
    }
    //+------------------------------------------------------------------+
    //| Perceptron Price |
    //+------------------------------------------------------------------+
    double PP(int i, int x)
    {
    double R1 = (Close[i] + Open[i] + High[i] + Low[i]) / 4;
    double R2 = (Close[x] + Open[x] + High[x] + Low[x]) / 4;
    double R = (R1 - R2) / Point / 10;

    return(R);
    }

  6. #6

    Default

    Quote Originally Posted by doshur View Post
    Hi,

    I myself had created a EA with this perceptron filter and was working well also.

    I agree it is time consuming and would like to find out more and learn more about this perceptron thing......
    }
    Sorry I am not a coder, I can't relate to that. May be Funyo can.
    I learn a lot on how to optimize from testing this system and reading threads.
    MQL is still a foreign language to me. I learn when I have time.
    ****
    P.S. You should learn Russian. They love high intellectual chats with a cup of tea. It's a part of their social ceremony.
    Last edited by Metatrader7; 08-02-2009 at 06:19.

  7. #7
    Junior Member
    Join Date
    Jun 2009
    Posts
    22

    Default

    Hi Metatrader7:

    Thanks for your Information and optimize files, I backtested it and optimized it on buyX.set file and only go long, it had good results, I might be wrong, it seems better than those Indicators base EA.

    If I optimize on Buyx.set file and only go long, After optimization, Which setting should I pick?
    - most profit?
    - mix drawdown?
    - Max payoff? ETC

    I am using M5, Is it scraping EA work well on shorter time frame only?

  8. #8

    Default

    I have to get back on that later this week. I spent too much time with EAs, I have to keep up with my manual setup. I have to go into Russian chat room and check their posts again. They figured an entire procedure to optimize it so it produces good results on forward tests. That's what they are working at - live results. There are lots of posts on this project. Russians are way ahead of any English speaking forum. They have their marketing junk too, but there are many high level pro posts and lots of things going. They are much serious about trading biz overall.
    ***
    Another area - they understand importance of broker data collection and maintenance. They have few scipts to pull data from server, and to patch data gaps. I have to test these scripts as well to see how that works. There is lots of stuff coming. This EA needs quality data to optimize correctly.
    ********
    This isn't a scalper, it cuts 100-200 pips (average) chunks out of euro.

  9. #9
    Junior Member
    Join Date
    Mar 2009
    Posts
    6

    Default

    Pls help us with the training of the perceptrons the more accurate way

    Thanks

  10. #10
    Junior Member
    Join Date
    Jun 2009
    Posts
    22

    Default

    Thanks Metatrader7 for your detail real world infomation.

    you are right, This EA takes so much time for optimization.

    It took 24 hours doing buy1.set for only recently 7 months data.

    My plan is that, After Level 1 primary optimization for recently 7 months, I would like to run it for last year data, so I might be able to tell that this EA works or not. ( Data that they haven't optimized)

    Would you tell me that how I can combine Buy1.set with Sell2.set data?

    There are few versions of N7S EA here, Older version seems working for 4 digits broker and newer one for 5 digits.

    I am backtesting on older version/5 digits of N7s for now, I find that older set file is not working on your newer N7S with filters.

    I tested many simple indicators base EAs, It seems that most of them don't work well or too danger to run, or they look so well on backtests but run so poor in forward tests, or Some EAs run well forward test by hiding lose using un-close Payload but suddenly blow all money.

    Sorry for my poor English, I am not a programmer so I don't know them very well, I builded a Tiny Z80 computer before so I learned
    some Assemble languge, I find that MT4 programming is not that hard.

    I also find that good programming skill help but it is not a key for a winning EA but good Idea of good EAs is.

Tags for this Thread

100, advisor, average, backtest, backtesting, broker, coding, demo, ea works, eas, expert, expert advisor, experts, free, hedge, high, how to, indicator, indicators, intraday, learn, live, live account, low, main, manual, manual trading, martingale, metatrader, mq4, mql, mql4, mt4, profit, profitable, programmer, programming, real, scalper, scalping, scripts, server, simple, stop, study, system, test, time, trader, trading, training

Posting Permissions

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