stevehopwoodforex.com
https://www.stevehopwoodforex.com/phpBB3/
Print view

Can someone help make a EA?
https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=273		 Page 2 of 4
Author:  FX Juju [ Mon Jan 23, 2012 11:31 am ]
Post subject:  Re: Can someone help make a EA?
For now, the EA does not check if the TP/SL values are allowed by your broker... So if you set them too low, it will generate "errors 130"... I ll add that next time...
Author:  johnny2pips [ Mon Jan 23, 2012 3:41 pm ]
Post subject:  Re: Can someone help make a EA?
FX Juju wrote:Ok Johnny, I included user defined TP and SL for short and long positions... values set to 0 will be ignored... So now you can easily play with parameters... Also the TP and SL (if different from 0) are now set as soon that the order is opened...

One thing, I m not sure about CCI 236... Does "green" or "red" simply means it is above or below 0 ? If I remember well, in the double Woody CCI, the histogram display green and red, but also other colors near zero... Is there any special user defined level ?

I ll upload NZDUSD Ducascopy data and perform again a backtest after setting the Takeprofit to 0 (=no TP)...

After doing visual backtesting, you can easily add indies to the chart and see if the EA opens same positions as manual application of your strategy... Please check and let me know...
Awesome thanks again bud. I'll do some testing and let you know how it goes.
Author:  FX Juju [ Mon Jan 23, 2012 6:48 pm ]
Post subject:  Re: Can someone help make a EA?
Here is quick backtest (90% quality only)... but rather interesting for a first test...
NZDUSD, 1H, SL 25 pip, no TP, exit Heiken Ashi...
Author:  FX Juju [ Mon Jan 23, 2012 6:54 pm ]
Post subject:  Re: Can someone help make a EA?
Forgot to say... it s a one year backtest (2011)..
Author:  johnny2pips [ Mon Jan 23, 2012 9:09 pm ]
Post subject:  Re: Can someone help make a EA?
FX Juju wrote:Here is quick backtest (90% quality only)... but rather interesting for a first test...
NZDUSD, 1H, SL 25 pip, no TP, exit Heiken Ashi...
Talk about beautiful test results. Sure after the start there was a drawdown and maybe because of that some people would disagree. I'm my opinion it looks promising. This only improves my confidence in the system. Really appreciate your help here and am looking forward to further improving the system if at all possible.

Now that we have a fully mechanical back test posted what are your (anyone) thoughts on the results?
Author:  markft [ Mon Jan 23, 2012 11:58 pm ]
Post subject:  Re: Can someone help make a EA?
johnny2pips wrote:
FX Juju wrote:Here is quick backtest (90% quality only)... but rather interesting for a first test...
NZDUSD, 1H, SL 25 pip, no TP, exit Heiken Ashi...
Talk about beautiful test results. Sure after the start there was a drawdown and maybe because of that some people would disagree. I'm my opinion it looks promising. This only improves my confidence in the system. Really appreciate your help here and am looking forward to further improving the system if at all possible.

Now that we have a fully mechanical back test posted what are your (anyone) thoughts on the results?
The back test looks beautiful. In line with cut your losses and let winners run principle... :)
Author:  FX Juju [ Tue Jan 24, 2012 3:48 am ]
Post subject:  Re: Can someone help make a EA?
On my side, I also think it is promising enough to spend more time on the code :-)
I'm not very good at improving strategies... but...

Cutting losses and letting winners run as reminds Markft makes sense of course... that's a first thing...
I'm not sure a trailing stop will make it, but we can think of setting up kinda jumping SL that saves your profit once reached some levels...

I don't think if it will be enough though... I haven't carefully checked the backtests results yet, but at a first glance, it seems that while many trades (winners) are placed in the trend, some orders (the loosers) are placed in weaks trends or ends of trends... So maybe an additional filter may help...

It would be pretty easy to add multitimeframe double CCI... It would be a bit harder to add multitimeframe smoothed Heiken Ashi (due to the use of arrays), but possible (a guy did so for the Invictus system on FF) and I have the feeling (I might be wrong) that it would work better than double CCI to filter weak trend...

Average Directionnal Movement Index (ADX), checking MACD or MA convergence/divergence, or other indies might be other alternatives... Before adding anything, a visual examination of the backtest chart might help clarify where the strategy fails...
Author:  johnny2pips [ Tue Jan 24, 2012 5:30 pm ]
Post subject:  Re: Can someone help make a EA?
I think the problem is mostly due to trades being entered in low volume ranges and/or low volume trading hours. This is just a hunch as I haven't fully looked into this but maybe if we were to limit the trading hours from 12am to 3pm (again rough estimate) after 3pm the ea would only manage open trade and not open any more.

One thing I have looked into is major trend start times (100+ pip moves). From a few hours before london open to mid new/late york. This may or may not help at all.

As far as looking for more/better filtering indicators, multi timeframe indicators, I'm not sure how useful that would be because the higher time frames would repaint the current bar (just like normal has does) it would be interesting to see if it improves results.

Will look into further improvements.
Maybe we should move this talk to a new thread so people can know all the system details and if willing help with improvements?
Author:  johnny2pips [ Wed Jan 25, 2012 4:10 am ]
Post subject:  Re: Can someone help make a EA?
Quick couple of questions for ya.

If I don't check CCI 61 will the system trade based off has and CCI 236?
How long should it take to run a years worth of back test with the every tick option? I've had one running for over 24 hours and it's not even half done.
Author:  FX Juju [ Wed Jan 25, 2012 10:36 am ]
Post subject:  Re: Can someone help make a EA?
Hi Johnny,

The backtest I posted last time was already the one which is called "every tick" test in Empty4... That means you are basically running the same test as me... I just used Ducascopy data instead of broker data... I don t understand why your test takes so long... from my laptop it takes 1 minute !! That s just crazy...

The only way to improve that backtest was to used real tick data using a method described here: http://eareview.net/tick-data

So did I and here are the updated results... 99% modelling quality this time

You will see that the performance of the strategy slightly decreases but not that much. The equity curve is very similar...
All times are UTC Page 2 of 4

Powered by phpBB® Forum Software © phpBB Limited