Announcement

Collapse
No announcement yet.

AI4Forex EA - Free to download and test - EA attached

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • AI4Forex EA - Free to download and test - EA attached

    Hi everyone. I'm Walter from Italy.
    I'm a customer of forexsignals since many year. I tried some independent signals provider and some sponsored one like TrustedFx and than Viper, NightFox, Kilimanjaro, Goldstar, SteadyCapture. Sometimes a made some profit and sometimes it was a nightmare.
    In these years I tried to learn something from my experience and I started to lower my leverage (from 3x to 1x of master), select less provider at one time (from 5 to 1 or 2). Now I lose much less money, but I lose anyway.
    So I decided to spend a lot of time cosing my own expert (I'm a very good programmer but a bad trader...).
    In my experience I've seen many many technique but sooner ar later all failed.
    My conclusion is that only math and brute force can hack the forex market. Yes, for me the solution is the devil...grid+murtingale.
    I know, I know. ALL martingale will crash, it's just matter of time. I totally agree.
    This experiment shows how itís possible to create a grid that doesnít crash since 2000, passing 2008 and 2014, the most volatile years, the beast for any grids.
    This is possible because every day the grid analize what I call DNA of the instrument and adapt itself to current market situation. I know that this is not something new, but it seems to work for me.

    The aim of this post is to cooperate to create a usefull EA and if it works and someone is interested in, I will create a signals on this site and monetize it.

    My Backtest Settings:
    Alpani NZ MT4
    History downloaded from History center
    Pair: EURUSD
    TimeFrame: M1 (itís selected by the logic)
    Spread: 10
    Starting Balance: 500.000$
    TesterGraph.gif
    report.png
    Attached Files
    Last edited by IA4Forex; 03-12-2018, 04:25 PM.

  • #2
    MYFXBook Strategy test
    I had to split the report in different periods because has a limitation of 50mb per report but all periods have been tested with the same settings.
    This is just a starting point, I know that the return isn't worth the risk.
    For the moment with this first expert I want to show how is it possible to have a grid suitable for almost 20 years of trading.
    Last edited by IA4Forex; 03-09-2018, 09:55 AM.

    Comment


    • #3
      Optimized and aggressive compounding settings:
      AI4Forex EURUSD 2000-2018 Extreme Settings.png

      Comment


      • #4
        This Expert uses Artificial Intelligence to make a simple "Moving Average" expert advisor into a winning long term system.
        This is the chart tested from 2000 to 2018. It open a trade at open price and close it at the close of the same bar. That's why I used "Control points" method.
        EURUSD H4
        AI4Forex Moving Average.gif
        Backtest is very slow because if the machine learning that is launched every bar.
        Attached Files

        Comment


        • #5
          Originally posted by RenkoGuy

          Can you make a version for MT 5 ?
          Yes, sure. I'll attach here as soon as I will have coded it.

          Comment


          • #6
            Originally posted by RenkoGuy

            Very nice !
            I added the MT5 version to the first post.

            Comment


            • #7
              Sure:

              input bool LongActive=true; Enable Long Baskets
              input bool ShortActive=true; Enable Short Baskets
              input int LongMagic=1; Magic number for long positions
              input int ShortMagic=2; Magic number for short positions
              input string Commento="AI4Forex"; Trade Comment
              input int DynamicStepFactor=100; Calculated Step between levels of the grid is diveded per this value
              input int DynamicDeviationFactor=3; The standard deviation is moltiplied per this value
              input int DynamicProfitFactor=1; The profic factor is moltiplied for this value
              input double BaseLot=0.01; The base Lot
              input int LotEvery=50000; a base lot is opened every "Lotevery". So if you inter 50.000 and balance is 500.000, the base lot wil be 500.000/50.000=0.10
              input double LotFactor=1.1; The tot factor in the martingale
              input int LotDecimal=2; The decimal used in lot calculation


              if you get base lot =0 maybe the balance is too low

              Comment


              • #8
                Modelling quality - 25% ... Real results won't even be close to the backtest results

                Comment


                • #9
                  Originally posted by Amadorian View Post
                  Modelling quality - 25% ... Real results won't even be close to the backtest results
                  That's a limitation of MT4 on M1 bar: https://www.mql5.com/en/articles/1513
                  Here it is the backtest in mt5 with "Real Tick data", that is base on real tick, not the same mt4 tick modeling.
                  So, real spred, real tick on news and so on...
                  Attached Files

                  Comment


                  • #10
                    Use tickstory for a 99% modeling quality backtest. 25% is pretty much worthless.

                    Comment


                    • #11
                      Originally posted by Amadorian View Post
                      Use tickstory for a 99% modeling quality backtest. 25% is pretty much worthless.
                      MT5 stats has be done with "Real tick". There is nothing better that this. https://www.mql5.com/en/articles/2612

                      Comment


                      • #12
                        As long as your modeling quality is below 99%, there is something to improve. Do the backtest with tickstory and MT4, i would be very surprised if the results are even close to what you get now.
                        After all it will safe you money doint it the right way.

                        Comment


                        • #13
                          Originally posted by Amadorian View Post
                          As long as your modeling quality is below 99%, there is something to improve. Do the backtest with tickstory and MT4, i would be very surprised if the results are even close to what you get now.
                          After all it will safe you money doint it the right way.
                          I give up!

                          Comment


                          • #14
                            This is the backtest of a boosted version from 2013 to 2018.
                            Modeling quality is 100%, using real tick data.
                            1370% with max DD=28%
                            Attached Files

                            Comment


                            • #15
                              Now we can see a problem, the margin level on this backtest went down to only 83%. That is extremely close to a margin call on some brokers. (e.g. VantageFX) and many other brokers require a 100% margn.
                              Last edited by Amadorian; 03-13-2018, 02:38 PM.

                              Comment

                              Working...
                              X