QuantDean
Junior Member
Posts: 6
Registered: 9-5-2017
Member Is Offline
Mood: No Mood
|
|
Matching Results between TS and GSB reports
I imported some data (ES.D 30min) from TS to GSB, built some strategies using just ES data, however I cant get the equity curves or results to match
between TS and GSB. They are very different. TS is consistently showing less performance in all the key metrics.
What would be the possible reasons for this?Â
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
The most common issue is the session times are different. GSB ES.x data uses 830 to 1500, not 830 to 1515.
You could also try the data that came with GSB to see if the same issue occurs.
So use the ES.30 with no secondary data streams. Export the system to TS with 830 to 1500 session time.
I know your problem is severe, but it is common for metrics to be 99% the same. This is totally fine, and not worth pursuing.
The reasons are bars back setting, different mathematical degrees in precision between GSB and TS, and TS data varies from GSB data but some amounts.
Please keep us informed.
|
|
|
QuantDean
Junior Member
Posts: 6
Registered: 9-5-2017
Member Is Offline
Mood: No Mood
|
|
Initially I exported the data from TS without adjusting sessions. Assuming the sessions are not adjusted in GSB, the results should match,
Do you mean that GSB will override the sessions from TS export and always use 830 to 1500?
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
GSB will not override the settings. Do you ahve 0$ slippage & commission on GSB and TS?
Worst case you can put debug script on, and nail exactly whats different.
You could also try if all GSB systems are different, or just some.
|
|
|
gmoney
Junior Member
Posts: 14
Registered: 8-5-2017
Member Is Offline
Mood: No Mood
|
|
With GSB 22.0, results between GSB and TS are very different. In some cases within 25% difference, and sometimes wildly different (nice profit in GSB,
losses in TS). I triple checked commission, slippage, custom sessions etc. In fact, I imported the exact same data files into TS for comparison and
results are still different. This is true for every system and symbol I tested.
I tested 20.4 and results between GSB and TS match almost perfectly.
I have screenshots and the code if you want me to post those. How do I embed or attach those in a forum post?
HERE ARE SCREENSHOTS
 
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
You can do a post, then edit and attache files.
Upload on the forum is very slow. If files are large .7zip works better than zip. Either should be ok.
You should be on 23.2 build
You need to have diag on so the results of each indicator can be compared from gsb to ts.
I expect a number of bug fixes this week. We need to figure out of the differences are due to a specific indicator or not.
|
|
|
gmoney
Junior Member
Posts: 14
Registered: 8-5-2017
Member Is Offline
Mood: No Mood
|
|
Screenshots added. If you want the code also, please let me know.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Youve got indicators which have been in GSB for some time. I was expecting problems with the newer indicator only.
The ability to put a known system from GSB to another was removed when we had secondary filters added. It will come back in.
So I cant reproduce this in my GSB.
You are going to have to put diag mode on, click off the system then back on to the system.
Then the diag info will be in the TS code.
Put the GSBTS code in TS, and compare the print log.
Secondary filters diag is not in GSB till the next build
Then you or I need to compare the logs to see the issue.
Im hopping to make an automated version of this in GSB down the track.
I enclose screen shots of me doing this.
Its going to be common to get a difference in the beginning of data, but this can be ignored. Reason can be maxbarsback are different etc
You did a good job by using gsb txt data in TS. That's also a reason for differences which youve ruled out in this case
 
|
|
|
gmoney
Junior Member
Posts: 14
Registered: 8-5-2017
Member Is Offline
Mood: No Mood
|
|
Where in GSB 24.1 can I turn on the diag GSBTS code? Before it was called debug but it appears to have been removed.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
turn advanced mode on under app settings gui
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Todays build of GSB will have fixes for issues with the new indicators, and some of the old ones.
Also it will have secondary filters diag info in the debug file. SHould come out in the next 4 hours I hope.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
24.5 still has issues but its much improved. Im getting better performance in ts compared to gsb
I tested about 9 systems and two were out. Both ts versions were better than GSB
 
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I am planing to have a GSB feature where you dump the TS print log into GSB, and it diagnoses the issue.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I am often finding adjusting the variable sfEntryLevel seems to fix many of the miss-matches.
If result > entryLevel AND sfResult > sfEntryLevel And sfResult[1] > sfEntryLevel Then ..
Still waiting for the diag tool to be built
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I now have 6 of 6 systems matching. 100.0000% match to TS, and 98+ % match to MC. (often 2 trades out)
     
Attachment: Login to view the details
|
|
|
![[*]](images/davis/default_icon.gif){kind=icon}
