Expert Advisor Strategy | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

lets share / create expert advisor strategy from indicator 25 replies

Super Carry Trading Strategy - Expert Advisor 22 replies

Expert Advisor WIth fibonacci Exit Strategy 0 replies

Expert Advisor 8 replies

Interbank FX Expert Advisor how to program? 15 replies

Attachments: Expert Advisor Strategy
Exit Attachments

Expert Advisor Strategy

  • #41
  • Apr 28, 2007 9:28am Apr 28, 2007 9:28am
Wow - very impressive - broke the 1,000,000 mark.
Attached Image (click to enlarge)
Click to Enlarge Name: results.jpg Size: 24 KB
  • #42
  • Apr 28, 2007 9:34am Apr 28, 2007 9:34am
Quoting alsup1955
I,m trying to copy these to my desktop to load. For some reason it won't copy as the yellow yeild sign. I have copied many before. Are these files good? any ideas?
Those are compiled EA and indicator (*.ex4).
The yield sign shows on open source EAs and indicators (*.mq4).
  • #43
  • Apr 28, 2007 9:54am Apr 28, 2007 9:54am
Thanks aqibsherwani, very interesting concept. I'll begin forward testing this week. I like the idea of having an EA running along side of my live trading with a portion of an additional account set aside for EA use.

Cheers,
Thom
  • #44
  • Apr 28, 2007 10:46am Apr 28, 2007 10:46am
Quoting aqibsherwani
The EA is optimised for Eur/Jpy. Please try on 1Hr time frame with my settings as mentioned in the first post. Or you can try your own settings.
Ok Aqib, this optimisation apparently works for EURJPY, but since we have no further info from you on the algorithm/methodology I cannot really say if it is luck with that pair or not. For example, minor changes (0.5)to the buy level will take you from 22% drawdown to 85% drawdown. Also, the number of trades over 3 years was 150 using the H1 bars as you said. That is extremely low and I personally would only use an indicator for this kind of frequency, not an EA to trade on my behalf.

Please provide either a description of what you are doing, what these "buy" and "sell" levels mean or at least alternative settings to use with other pairs.

Thanks

P.S: I am a computer scientist as well, so don't worry about overly technical descriptions
Virtue finds and chooses the mean. Aristotle, Ethica Nichomachea
  • #45
  • Apr 28, 2007 11:03am Apr 28, 2007 11:03am
Quoting Plutonite
Ok Aqib, this optimisation apparently works for EURJPY, but since we have no further info from you on the algorithm/methodology I cannot really say if it is luck with that pair or not. For example, minor changes (0.5)to the buy level will take you from 22% drawdown to 85% drawdown. Also, the number of trades over 3 years was 150 using the H1 bars as you said. That is extremely low and I personally would only use an indicator for this kind of frequency, not an EA to trade on my behalf.

Please provide either a description of what you are doing, what these "buy" and "sell" levels mean or at least alternative settings to use with other pairs.

Thanks

P.S: I am a computer scientist as well, so don't worry about overly technical descriptions
You are getting low trades because only trades with high probability are executed. The buy level is that you are taking trades with 0.7 (70%) are likely to be profitable.

This EA is designed for eur/jpy and is no luck it is performing good.
  • #46
  • Apr 28, 2007 11:16am Apr 28, 2007 11:16am
Quoting Plutonite
I have the same problem. It is not being drawn. EA works but results are quite terrible I am afraid with 90% accurate data on H1 charts GBPJPY and EURUSD.

In short, maximal drawdown was 90.% and profitable trades (total) was 13%. Default values used (risk = 10). Will try EURJPY as well, although /JPY crosses have similar action.
if this system 90% fails on G/Y
then maybe we can do the opposite on G/Y
so it will profit 90%

just a dummy thought....
ha ha ha ha ha

btw thanx for sharing this system aqibsherwani

I'll give it a try this week
Every moment in the market is unique
  • #47
  • Apr 28, 2007 11:16am Apr 28, 2007 11:16am
Do realize though that evolutionary algorithms are used to find optimal parameters for an algorithm. It basically achieves the same results as brute forcing, mostly just a little worse, but much faster. That's the only thing evolutionary (genetic) algorithms mean - optimization of values.

So basically this is curve-fitting. If it's a good system, that doesn't matter, but it is curve-fitting.
  • #48
  • Apr 28, 2007 11:56am Apr 28, 2007 11:56am
Quoting aqibsherwani
You are getting low trades because only trades with high probability are executed. The buy level is that you are taking trades with 0.7 (70%) are likely to be profitable.

This EA is designed for eur/jpy and is no luck it is performing good.
Thanks for the reply,

But the % of profitable trades is unfortunately less than 50% even with the EURJPY alone. The number of consecutive losses is more than twice that of wins..And when you increase the Buy_level to .8 it goes down, and drawdown even goes as far as 84% for .85!! Maybe the algorithm needs some polishing in this estimation business?

Also, if changing the parameters (stop loss..etc) is not sufficient to adapt the algorithm to GBPJPY, which is almost IDENTICAL in price action/movements, then your AI is definitely very complicated indeed. I notice also that the EA is most profitable in 2006-7, and after seeing/wrinting hundreds of EA's I can tell you this usually means that this period's curve fit much better the others, which is very dangerous.

Anyway please let us know if you decide to share the indicator or the algorithm, and good luck!
Virtue finds and chooses the mean. Aristotle, Ethica Nichomachea
  • #49
  • Apr 28, 2007 12:12pm Apr 28, 2007 12:12pm
Quoting Plutonite
Thanks for the reply,

But the % of profitable trades is unfortunately less than 50% even with the EURJPY alone. The number of consecutive losses is more than twice that of wins..And when you increase the Buy_level to .8 it goes down, and drawdown even goes as far as 84% for .85!! Maybe the algorithm needs some polishing in this estimation business?

Also, if changing the parameters (stop loss..etc) is not sufficient to adapt the algorithm to GBPJPY, which is almost IDENTICAL in price action/movements, then your AI is definitely very complicated indeed. I notice also that the EA is most profitable in 2006-7, and after seeing/wrinting hundreds of EA's I can tell you this usually means that this period's curve fit much better the others, which is very dangerous.

Anyway please let us know if you decide to share the indicator or the algorithm, and good luck!
GBP/JPY works for me, at least in backtesting. USD/JPY also works, although it only produces a small profit.
  • #50
  • Apr 28, 2007 12:22pm Apr 28, 2007 12:22pm
Quoting ThomasB
Do realize though that evolutionary algorithms are used to find optimal parameters for an algorithm. It basically achieves the same results as brute forcing, mostly just a little worse, but much faster. That's the only thing evolutionary (genetic) algorithms mean - optimization of values.

So basically this is curve-fitting. If it's a good system, that doesn't matter, but it is curve-fitting.


That was my impression of evolutionary algorithms too.....
  • #51
  • Apr 28, 2007 1:08pm Apr 28, 2007 1:08pm
Quoting aqibsherwani
i have just started forward testing and it is giving very good results. My backtest results have been very good too. But I dont rely on backtest results. You can backtest yourelf to get an idea.
"Only if the mutant has a higher fitness than the parent does it become the parent of the next generation. Otherwise the mutant is disregarded." Does this mean it as always re-writing itself to become better based on what it learns? Self optimisation?
Do not focus on making money; focus on protecting what you have.
  • #52
  • Apr 29, 2007 12:30am Apr 29, 2007 12:30am
What I don't understand is why on a same period of time, if risk set to 20 the number of trades is higher (23 for eurjpy and 54 for gbpjpy for the period considered in my backtest) while when I set the risk to 5, there is only one trade for each of them. Shouldn't it take the same entries?
  • #53
  • Apr 29, 2007 1:34am Apr 29, 2007 1:34am
>it's genetic auto-optimization.<

Oh my there is breeding going on in forex and I missed that lesson.

guess Ill just stay dumb on this one.
  • #54
  • May 1, 2007 10:29am May 1, 2007 10:29am
So I'm forward testing this EA (default settings) on a demo account. It took its first trade yesterday @ 18:00 GMT long on the EUR/JPY. It bought 11 lots @ 163.06 with a S/L @ 162.56. The trade is currently still open reached a high of 163.28 (05/01/07 03:00GMT) and is now sitting at a loss @ 162.87. I think an optimization with a T/P would be useful. Any plans for that in the future aqibsherwani? I know you're already planning on a trailing stop feature which would be nice.
  • #55
  • May 1, 2007 3:20pm May 1, 2007 3:20pm
Thanks for the work you have done on the indicator and EA. I have dedicated a computer to running only your indicator and ea. Since Sunday I am up $437. on my demo account using GBP/JPY. I have it set also for EUR/GBP and USD/JPY but it hasn't taken a trade on either as of yet. I have my buy level set to .5 and risk at 20, everything else is set at default. Do you plan on allowing users to see the indicator at work any time in the future? Thanks again.

Mr4x
  • #56
  • May 2, 2007 12:23pm May 2, 2007 12:23pm
Quoting MasterOfDomain
So I'm forward testing this EA (default settings) on a demo account. It took its first trade yesterday @ 18:00 GMT long on the EUR/JPY. It bought 11 lots @ 163.06 with a S/L @ 162.56. The trade is currently still open reached a high of 163.28 (05/01/07 03:00GMT) and is now sitting at a loss @ 162.87. I think an optimization with a T/P would be useful. Any plans for that in the future aqibsherwani? I know you're already planning on a trailing stop feature which would be nice.
hi i have implemented a trailing stop feature which i let release soon.
  • #57
  • May 2, 2007 12:23pm May 2, 2007 12:23pm
Quoting mr4x
Thanks for the work you have done on the indicator and EA. I have dedicated a computer to running only your indicator and ea. Since Sunday I am up $437. on my demo account using GBP/JPY. I have it set also for EUR/GBP and USD/JPY but it hasn't taken a trade on either as of yet. I have my buy level set to .5 and risk at 20, everything else is set at default. Do you plan on allowing users to see the indicator at work any time in the future? Thanks again.

Mr4x
at the moment i am developing the EA and i belive it can earn you alot. i will release the ea and ind. source code if i find this ea is useless just like the others.
  • #58
  • May 4, 2007 6:56am May 4, 2007 6:56am
Ok, so im trying out the EA and it initiated a buy a couple of hours ago.
The funny thing is that Take Profit is set to zero. This can't be normal can it?

See attached screen shot


[IMG]file:///C:/DOCUME%7E1/rnystran/LOCALS%7E1/Temp/moz-screenshot.jpg[/IMG]
Attached Image (click to enlarge)
Click to Enlarge Name: quibsterEA.jpg Size: 5 KB
Liberate me ex inferis
  • #59
  • May 4, 2007 7:09am May 4, 2007 7:09am
Quoting Ronyn76
Ok, so im trying out the EA and it initiated a buy a couple of hours ago.
The funny thing is that Take Profit is set to zero. This can't be normal can it?

See attached screen shot


[IMG]file:///C:/DOCUME%7E1/rnystran/LOCALS%7E1/Temp/moz-screenshot.jpg[/IMG]

There is no built in T/P setting. You have to set it manually after it initiates an order.

BTW, you can't attach images that are on your computer locally to appear online.
  • #60
  • May 4, 2007 7:52am May 4, 2007 7:52am
could please post someone a more or less reliable backtest with 90% modelling quality - all others are useless.
0 traders viewing now