Complete testing instructions and EA files
Here what you asked for, please post your results here.
Author estimate live potential - 10% monthly with 20% DD on average.
But this EA is 100% optimization reliant, you might beat it as well.
Please post entire cycle tests from optimization back test to resulting forward with the weekly score per currency and then live. That's how it was done on MQ4 forum.
Letest test version takes only 20 min per optimization cycle.
Last edited by Metatrader7; 08-09-2009 at 23:13.
Please read the Instruction first, many answers are there, see previous post
Thanks so much for your great help.
I did play with L3 version 2 days ago
I optimized it using 5 min data from my demo provider FXpro for 8 months, Stage 1 only buy + sell .set took me 45 hours and it turned 100,000 to about $138,000 in 8 months with about 60+% winning.. most of them are trending winning rather than scalping which mean this EA is broker friendly EA ( broker don't hate it)
Then I downloaded data from from Mt4 history Metaquotes sever for longer range backtest, it screwed up my data( May be GMT time problem, When I retested same 8 months which was profits $38,000 , changed to loss $9,000
I re-downloaded data from Fxpro, It gave me 5 months history data only, My profits become $10,000, it is much lower then before.
I am not sure why?, I check the latest data, It seems that it is much more rough, I might be wrong that sometime data from demo server is smoother than live server, For Indicator base EAs, It is not a problem, But not for N7S precetron EA, They are pretty sensitive to data quality.
I don't have live account, I will open a live account from FXpro then I can download live data for optimization. It might take me sometime to do it.
There are many interesting EAs in this site, Just don't have enough time to play with them, That Taichi Demarker EA run pretty good, I can not beleive that it is Free, Also, this is a well coded EA and very easy to modify, It seems that it still have room to adjust with trigger coding section for better/accurate results, Also another nice program Pseudocarcharias, It seems pretty good too, Thanks for those coders for their great ideas, Also I really like Funyoo coding style and speed, I also download somewhere a hackable demo version Robominer 2.1 , It run so well in backtest, However, when I run it in demo, It is ok but not as good as backtest, it seems that it is able to predict price movement and put down 40pips profit target but price only move to about 30 pips, However, it is still amazing.
Optimization really taking time, I need 2 more computer, I bought 2 used computer for $100, I am trying to buy a KVM keyboard/mouse/monitor switch, so I can run optimization on other computers and easy to switching between them.
Last edited by Ricky101; 08-10-2009 at 10:43.
L3 is an old version, that's why it takes so long. Only after L7 they fixed all bugs and made it Live.
Try later versions and re read instruction
Check newly uploaded version
This is the version specifically designed to run optimizations at a fraction of time, takes only 20 min to optimize.
then you load results into N7S_AO_772012_M5b.mq4 to run on charts.
Thanks for you doing it.
I download all files and I follow instruction carefully, It seems missing information on Optimize M5_test version.
However, I am using enclosed (4 weeks plus 2 weeks Optimization) L7 version files and method, I tried it on N7S_AO_772012_M5b_test.mq4 , I think that there are some information missing, it took me total 30 hours to finish with pretty good profit, Then I back test it with N7s_M5B version for 5 months, I am very surprise that results are so bad, I end up losing about 20% money, I am not sure why?
I am not able to open up live account with FXpro due to passport issue so I don't have live account with live server's data for testing, Demo servers are not reliable, I download data's from different Fxpro's demo servers, I compared it and they are not the same.
I am glad to hear that you are following steps. That you try to gain more control. EAs running is a hard work. Some think it's easier then manual trading. Who thinks that way simply have no idea. It's not simple, it's simply different with more rules and procedures to follow.
Originally Posted by Ricky101
I heard if system doesn't optimise to be profitable at certain week then don't run it on that currency.
Some tried different longer periods - 6 months with good results, but it increases optimization time and you have to keep data.
Also try to patch data, there is a script to pull missing ticks from broker server or to reconstruct them. keeping Data is your biz, broker has no interest in that, actually that's against their interest.
Latest version posted is M5b. L7 testing sets are compatible. Instruction is written for L3 as example, but compatible with later versions.
Last edited by Metatrader7; 08-13-2009 at 19:33.
Currency Market is always changing in both trending or ranging or combination of them.
Just guessing, N7S coding and "4 weeks plus latest 2 weeks" optimization are weighted on 2 weeks, because that 2 weeks are the possible one closer to next week future market. This method is not suitable to back test many months as Indicators base EA but it provides the greatest accuracy to predict next week market which is what we want.
Last edited by Ricky101; 08-13-2009 at 20:49.
Yes, now you got the idea. Level 1 and Level 2 periods combination could vary. You can even try them be the same, various overlap or no overlap, greater length. last 6 month is optimal optimization data range for intraday EAs
Originally Posted by Ricky101
Thanks I'll study this type of NN soon I was trying to make EA depend on another type of NN but the problem the EA did not work on backtest The NN was very heavy and Mql4 slowly.