Download MetaTrader 5

Tester in new build 604 doesn't work correct

To add comments, please log in or register
Find out how to install MetaTrader on Mac OS
kl-mueller
168
kl-mueller 2014.02.08 21:21 

Hi,

the new tester in the build >600 seems to have a bug. Possibly the tick-Generation is not precise enough.

Here 2 screenshots:

Old build 509 is working correct, the stop has released when backtesting the FDAX, the spike on last Thursday.




Here the tester of build 604, the same EA / same strategy / same history / same settings: the stop did nit release (stop-level is indicated with the yellow line)




Is there a possibility to improve the backtest accurateness?

Thanks,
Klaus

Ubzen
5394
Ubzen 2014.02.08 22:26  
kla-mue:

Hi,

the new tester in the build >600 seems to have a bug. Possibly the tick-Generation is not precise enough.

Here 2 screenshots:

Old build 509 is working correct, the stop has released when backtesting the FDAX, the spike on last Thursday.

Here the tester of build 604, the same EA / same strategy / same history / same settings: the stop did nit release (stop-level is indicated with the yellow line)

Is there a possibility to improve the backtest accurateness?

Thanks,
Klaus

  • We're clearly blind-sided about how these tests are being performed.
  • Would you mind telling us a little more details about how its being performed.
  • What are the Spreads between test?
  • Is both platforms still connected to the broker?
  • If so, keep in mind its the Weekend,
  • What type of order within the picture [ Buy || Sell ] ?
  • Lets start there ... we may need to see the codes later.
  • It could be a bug because built 600 is still new ... but lets find out.

Alain Verleyen
Moderator
30731
Alain Verleyen 2014.02.08 23:01  
ubzen:
  • We're clearly blind-sided about how these tests are being performed.
  • Would you mind telling us a little more details about how its being performed.
  • What are the Spreads between test?
  • Is both platforms still connected to the broker?
  • If so, keep in mind its the Weekend,
  • What type of order within the picture [ Buy || Sell ] ?
  • Lets start there ... we may need to see the codes later.
  • It could be a bug because built 600 is still new ... but lets find out.

Will you write all your comments in this way now? Looks like a robot.
Ubzen
5394
Ubzen 2014.02.08 23:07  
angevoyageur: Will you write all your comments in this way now? Looks like a robot.

This is what developing and trading a robot for a while does to a guy.

  • Does FDAX run on Bid and Ask?
kl-mueller
168
kl-mueller 2014.02.09 12:51  

Hi,

FDAX is the german index-CFD and has rates in Ask and Bid, like forex-pairs.

Of course it's a trading-robot, otherwise i could not make a backtest. The .ex4 is done with the old Compiler, may be it makes a difference?

Spread in the backtest was fixed @15. Weekend is not the problem because the spread is fixed.
I don't know the spread during live-session, but this is not important for backtesting-results.

The enter-logic determined to open a market long-position at 8:10 AM with StopLoss @9045

The testing-conditions were 100% the same, I noticed the increased speed of the new build. When I did the backtest in visual mode and very slowly, it seemed to overleap the highest/lowest ticks of the bar.

Amazing: i also tried both builds (509 / 604) on 2 operating-system, window7 and XP. With Windows XP both builds succeeded and the StopLoss released....!?


regards,

Klaus

aaaacom
117
aaaacom 2014.02.09 13:02  

hi Klaus, if you look "mismatched chart errors" in build 600, there are many bar data errors. causing tester fails.

bar data consistency is poor with build 600, yet.

Alain Verleyen
Moderator
30731
Alain Verleyen 2014.02.09 13:05  
cbalta:

hi Klaus, if you look "mismatched chart errors" in build 600, there are many bar data errors. causing tester fails.

bar data consistency is poor with build 600, yet.

Current build is 604, not 600.

Can you provides some clues that this is well related to the new build. If it's a bug, it has to be reproduced to be corrected. Thanks.

aaaacom
117
aaaacom 2014.02.09 13:26  

hi, actually, developers need to test, if product is working compared to previous stable one (509)..

developers should compare EA starategy tests of build 509 and build 604, before end user.


i'm trying to compare 3 test results from one EA:

1. 509 build ex4 under 604 (match with 2)

2. 604 build ex4 under 604 (match with1)

3. 509 build ex4 under 509 (this is what i trust currently.)


604 and 509 test do not match.. currently i rely on 509.


strategy tester posssibly lost too much details, compared to 509. but mq developers need to convert old mq4 code to new

mq4 and compare the test results of 509 and 604. thay do not match yet.


it may not be true, but new MT4 possibly will not run the same code, as it run under 509 on live. like tester does not yet.

Alain Verleyen
Moderator
30731
Alain Verleyen 2014.02.09 14:46  
cbalta:

hi, actually, developers need to test, if product is working compared to previous stable one (509)..

developers should compare EA starategy tests of build 509 and build 604, before end user.


i'm trying to compare 3 test results from one EA:

1. 509 build ex4 under 604 (match with 2)

2. 604 build ex4 under 604 (match with1)

3. 509 build ex4 under 509 (this is what i trust currently.)


604 and 509 test do not match.. currently i rely on 509.


strategy tester posssibly lost too much details, compared to 509. but mq developers need to convert old mq4 code to new

mq4 and compare the test results of 509 and 604. thay do not match yet.


it may not be true, but new MT4 possibly will not run the same code, as it run under 509 on live. like tester does not yet.


Ok, as you wish. If you think developers will resolve your problem without you giving any useful information, you are risking to wait a long time.
Simon Gniadkowski
Moderator
18001
Simon Gniadkowski 2014.02.09 14:58  
cbalta:

hi, actually, developers need to test, if product is working compared to previous stable one (509)..

developers should compare EA starategy tests of build 509 and build 604, before end user.


i'm trying to compare 3 test results from one EA:

1. 509 build ex4 under 604 (match with 2)

2. 604 build ex4 under 604 (match with1)

3. 509 build ex4 under 509 (this is what i trust currently.)


604 and 509 test do not match.. currently i rely on 509.

How are you getting the same History data into 509 ad 604 ?
aaaacom
117
aaaacom 2014.02.09 15:35  
angevoyageur:
Ok, as you wish. If you think developers will resolve your problem without you giving any useful information, you are risking to wait a long time.


developers, first, should test, whether the functions are healthy, compared to old version (509).

some test engineering required. since i have not in development, really hard to guess resources of problems.

here are what i doubt:

- bar data consistency. (i force tests on M1, if too many mismatched chart errors)

- precedence: compiler complains critical ones. but there my be some missed warnings.

-hedge: i saw a different behaviour of hedging, on 509 and 604, the same EA.

-iCustom, some of my external indicators are compiled withour error or warning, but not working. need to be fixed

-multiple order pieces may be problem.

-i've told how i compare test result: 3 test for one EA, comparing results. if you can setup a test environment with identical data, 509 ad 604, you can see the behaviours of an EA, on two different platforms.

for me, it is not easy to verify, 509 and 604 having the same data set. i'm just trying to make as match as possible.

sorry, but, but the upgrade is going very fast..

12345678...13
To add comments, please log in or register