MC 3.1 Beta optimization much slower than MC 3.0 release?

Questions about MultiCharts and user contributed studies.
glam_100
Posts: 157
Joined: 14 Jun 2006
Has thanked: 2 times
Been thanked: 9 times

MC 3.1 Beta optimization much slower than MC 3.0 release?

Postby glam_100 » 08 Jun 2008

Hi,

Somehow the optimization on MC 3.1 beta is much slower than 3.0 relase. Is this a beta thing or there is something wrong introduced to the new code?

I performed the "standard" TS speed test and got the following result:

Test symbol: SPY
Date: 12/31/2001 - 12/31/2003
Interval: Daily
Signal: Bollinger band LE (1..50, 1..5) and Bollinger band SE (1..50, 1..5)

result:

MC 3.0 release - 40sec
MC 3.1 beta - 1:40 sec

That's a huge difference.

glam_100
Posts: 157
Joined: 14 Jun 2006
Has thanked: 2 times
Been thanked: 9 times

Postby glam_100 » 11 Jun 2008

No response?

Marina?

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 11 Jun 2008

Hi glam_100,

We have tried to reproduced the problem on our end several times but the difference in optimization wasn't nearly as large as you're saying.

Please double-check that you have exactly the same data with the exact same number of trades and optimized parameters.

Also, did you run the same type of optimization in either case?

Regards.

glam_100
Posts: 157
Joined: 14 Jun 2006
Has thanked: 2 times
Been thanked: 9 times

Postby glam_100 » 11 Jun 2008

Hi Marina,


I just re-installed the new beta for re-testing. The result was very consistent - I got 40 sec with MC 3.0 release and 1:40 with MC 3.1 beta.

What I found is that when using MC 3.0 release the CPU load was close to 100% for all 4 cores while using MC 3.1 beta the CPU load was only 60%. So there are really some difference here.

The data I used is from TS data network:

symbol: SPY
date range: 12/31/2001 - 12/31/2003
strategy:
Bollinger Band LE ( length 1..50; numdevs 1..5 )
Bollinger Band SE ( length 1..50; numdevs 1..5 )
Total # of simulation: 62500

Results:

MC 3.0 release 40 secs
MC 3.1 beta 1:40 secs

Please find out what happened.

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 11 Jun 2008

Hi glam_100,

Could you please send us the workspace, the strategy, its settings, screenshots with optimization parameters for each of the runs? Also, if you ran the genetic optimization, we would need screenshots of the genetic optimization settings for each of the runs.

Regards.

glam_100
Posts: 157
Joined: 14 Jun 2006
Has thanked: 2 times
Been thanked: 9 times

Postby glam_100 » 11 Jun 2008

Marina I am not a QA personnel and I got work to do. I've already reinstalled MC 3.1 beta twice to reproduce the problem.

The problem is easy to reproduce. I suspect your staffs can't reproduce the problem because they have all the different versions of DLLs on their machine. Just find a clean machine, install the old MC 3.0 release and do a test run. Then install the new MC 3.1 beta and do another test run. You WILL see the difference. All the necessary information are on the post above (symbol, startegies, parameters, results).

This is a serious problem and I can reproduce it consistently.

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Postby Marina Pashkova » 12 Jun 2008

Marina I am not a QA personnel and I got work to do. I've already reinstalled MC 3.1 beta twice to reproduce the problem.

The problem is easy to reproduce. I suspect your staffs can't reproduce the problem because they have all the different versions of DLLs on their machine. Just find a clean machine, install the old MC 3.0 release and do a test run. Then install the new MC 3.1 beta and do another test run. You WILL see the difference. All the necessary information are on the post above (symbol, startegies, parameters, results).

This is a serious problem and I can reproduce it consistently.
Dear glam_100,

Our engineers reproduce dozens of scenarios every day and know how to make tests accurate. So far the situation you're describing hasn't been reproduced. Moreover, you're the only person who's reported this problem so far. To pursue this issue further, we would need more info from you (as requested above).

Regards.


Return to “MultiCharts”