Multi 10:4 Trader EA

[6y588]

I am using Dukascopy tick data to backtest. But looking at the code and iCustom, this may take a long while. I will start with YTD data on GU, but any reason why we use M5 for period?

[KevinT]

[MrLong]

Andy,

After being unable to display the data in v2.04a, I found your earlier v1.03 dash (see attached and screenshot.)

Looks like, regardless of screen or font sizing, I only get 24 not the full 28 pairs.
Any suggestions?

Regards,
Athar.

Hi,

Try YAD, that is the latest version that was modified by Candletiger: http://www.stevehopwoodforex.com/phpBB3 ... &hilit=YAD

Andy

Hi Andy,

You already adressed the subject (twice, afaik) but I can't find your last post on it:

-What to comment out and change on Multi 10:4 Trader EA code to use it as a 10.45 setup indicator only? (having Empty4 block live trading excluded!)

Thanks in advance

KevinT

No need to block anything from YAD, should run ootb.

[6y588]

Backtested GbpUsd, M5, Every Tick, 1Jan13 -10May13
-------------------------------------------------------
Bars in test 27535
Ticks modelled 5275211
Modelling quality 90%
Mismatched charts errors 0
Initial deposit 10000.00
Total trades 0

My results are different with no trades (using default settings).

1) If I want to test various pairs, for example do I set TestSymbol1=EURUSD?

[snailbeard]

Hello dear Brian,
very interesting your different backtesting. As there are so many variables in this EA and some of them are not "strict trading 10.4 rules" if activated, can you please post your set-file?

All the best to you and yours,
Dietmar[/quote]

Hi Dietmar,

I haven't used a set file, most of the settings are OOTB from the official version.
In the code directly I have changed:
useCSS=false

If we have enough volunteers we could share experimenting with different parameters,
but some of them would be pair dependent and unless we can run the tests over several years of data (as you know) we might be curve fitting anyway.

Not sure that I want to put much effort into back-testing and tweaking parameters as I have some derivative ideas for a new EA in my head which I would like to work on with modern s/w development tools...
I can't say that I enjoy working with Empty4 and MQL.
We could perhaps be much more productive in say Java/Netbeans and JForex and use the Empty4 bridge.

However, if enough people are interested I could produce a performance optimised version for faster back-testing.
It would be useful to have parameters specific to either each pair or group of pairs.
NZD pairs, for example, have often turned into losers for me, so there is an argument for more analysis.

The current settings are quite conservative, so draw-down is reduced but some trades are either missed or cut short by profit protection and some of those could be potential winners.

As you say there are so many possible tweaks it would be hard to find out which set works best when and on which pairs.

Are there enough people interested sharing the search for better configurations?
Do we have a shared repository of historical data?

Cheers

Brian

[snailbeard]

Backtested GbpUsd, M5, Every Tick, 1Jan13 -10May13
-------------------------------------------------------
Bars in test 27535
Ticks modelled 5275211
Modelling quality 90%
Mismatched charts errors 0
Initial deposit 10000.00
Total trades 0

My results are different with no trades (using default settings).

1) If I want to test various pairs, for example do I set TestSymbol1=EURUSD?

Yes that should work, with the same pair selected in StrategyTester.

M5 is a compromise because of the time it takes to do every tick,
M15 produced significantly different results
Every tick takes so much longer to run than M5.

[6y588]

I am getting zero divide during backtesting ...
a) Changing the tester period to see if that resolves the error (Note: changing the start date seemed to have solved it)
b) Using Dukascopy tick data from 2007 Apr to 2013 May
c) Historical data for several pairs UsdChf, EurUsd, AudUsd
d) My computer is sitting idle at the moment, so....

[forextrader-radioman]

[...]
I can't say that I enjoy working with Empty4 and MQL.
We could perhaps be much more productive in say Java/Netbeans and JForex and use the Empty4 bridge.
[...]
Cheers

Brian

Brian,
thank you for the reply.

I would prefer JForex as I have my real money account @ Dukascopy

, Dietmar

[snailbeard]

I am getting 100% CPU usage when the dash board is enabled sometimes
and lock up on the GP demo account, could be an infinite loop.

CPU usage is much lower with the dashboard disabled.

*** dangerous lot size ***

*** oversized lot size: sell stop on gbpcad on one account is 1.09 ***

Other accounts are giving lot sizes of 0.04, 0.20 for same sell stop

It is consistently repeatable on one account
Tried:
removing the expert, delete the pending order, then attaching the expert:
result: same order and lot size again
then
removing the expert, delete the pending order, restarting Empty4, then attaching the expert:
result:same order and lot size again
The minimum lot size on this account appears to be 0.10, could be anything to do with it?

[KevinT]