GSB Forums - Powered by XMB 1.9.11

There are 3 issues that I would like to report. I am using build 47.49.

First, the calculation of the number of data-streams that pass verification is mostly correct, but sometimes wrong. For example, it will report 1/3, when it should be 2/3.

Second, while testing NQ, it failed to copy all text files to the Worker folder on C (2 out of 16). I had to copy them manually. GSB helpfully issued an exception notification.

Third, it appears to me that the calculation of the average % deterioration is erroneous. This is so, under the assumption that the calculation is a simple average of the various metrics. I did this calculation several times and, in each case, there was an error. It could be that a weighted average is being used.

You should be on 48.16
1) Let me look via teamviewer or send support tick with screen shot of a system that is not correct.
2) The exception is likely file not found. Took us a long time to fix. Caused by identical file contents, different file name. Fixed in 48.16
3) This is not the formula used. Think its explained exactly in market validation video 2.
From memory its ave(fitnessA/FitnessB,NpA/NpB,AtA/AtV,Pfa/PfB

cyrus68 - 20-8-2018 at 12:55 AM

Yes indeed. File names are unique but contents may be similar. I just downloaded build 48.16.

I will send you an example of the verification error, next time I run GSB.

I will check the aforementioned video. The last line in my calculations in Excel show the % deterioration of each metric (B/A). For example, for Fitness, there is a 23.4% deterioration, and for NP, it is 13.5%. Then, a simple average is calculated for all the metrics. The number of trades is excluded for obvious reasons. The result is 11.3%, which differs from 15% calculated by GSB.

cyrus68 - 2-9-2018 at 10:40 PM

There appear to be significant problems with the SQL server. The Manager failed to aggregate systems, for ten minutes at a time. Then, suddenly, it would start to update systems produced by the workers. The net result is a waste of time and resources. If I want 3000 systems, I end up with 5000.

This isn’t the first time there have been problems with the SQL server. If it is frequently overloaded or suffers from other defects, the issue has to be resolved. A simpler solution is to have managers and workers communicate on a LAN. I suspect that the vast majority of users don’t use cloud workers or run GSB on remote computers.

Another issue is that pausing workers is no longer possible. They restart automatically after ten seconds.

admin - 2-9-2018 at 11:21 PM

There appear to be significant problems with the SQL server. The Manager failed to aggregate systems, for ten minutes at a time. Then, suddenly, it would start to update systems produced by the workers. The net result is a waste of time and resources. If I want 3000 systems, I end up with 5000.

This isn’t the first time there have been problems with the SQL server. If it is frequently overloaded or suffers from other defects, the issue has to be resolved. A simpler solution is to have managers and workers communicate on a LAN. I suspect that the vast majority of users don’t use cloud workers or run GSB on remote computers.

Another issue is that pausing workers is no longer possible. They restart automatically after ten seconds.

Do you mean no systems came until 10 minutes after the manager was started? How many minutes until the workers started?
If you want only 3000 systems, set the max unique systems to 3000 (under workplace left menu) and grace period to zero (under app settings ) and you will get 3000 systems.
I have the stats in sql data base, and the stats show the bulk of GSB users have there computers in data centers, not on lans. We will never go back to lan as it means we have sql and lan versions, and lan on anything but a local/nas drive is unreliable. ie dropbox / google drive dont cope.
You may have issues, but I have had zero. And i have been WF up to 180 systems at a time and running 90 workers the last week. (duie to the generosity of lots of GSB users) Today the SQL server load is near zero, so I doubt issues are in the sql server.
If you put GSB in admin mode, there are more stats I can work with to isolate your issue.
see this post.
https://trademaid.info/forum/viewthread.php?tid=92
You can then send a screen shot of the metrics
Shown here is 20118 systems, but workers produced 20588 systems. Why 400 systems were dropped is grace period was 30 seconds, so systems after 30 seconds were discarded. (There were about 90 workers on this run, and some of the workers are old slow machines)
22409-20588 were duplicate systems
Will look forward to this issue being resolved for you.

cyrus68 - 3-9-2018 at 04:49 AM

What I said is that the Manager would suddenly stop aggregating systems. In other words, the number of unique systems would remain constant. Ten minutes later, it would restart aggregating systems. In the meantime, thousands of systems had been produced by the workers. This happened repeatedly while I was running GSB on Saturday morning.

For example, the manager would stop aggregating at 2800. Ten minutes later, it would restart and add the 2000 in the queue. All I wanted was steady completion of the 3000, entered in the settings, so I could get on with testing another strategy, rather than waste time. If this wasn't an SQL server issue, then what was it?

Normally, I hope to be able to run GSB at my convenience, rather than when the SQL server is running smoothly. Hopefully, tomorrow will be a brighter day.

admin - 3-9-2018 at 04:05 PM

What I said is that the Manager would suddenly stop aggregating systems. In other words, the number of unique systems would remain constant. Ten minutes later, it would restart aggregating systems. In the meantime, thousands of systems had been produced by the workers. This happened repeatedly while I was running GSB on Saturday morning.

For example, the manager would stop aggregating at 2800. Ten minutes later, it would restart and add the 2000 in the queue. All I wanted was steady completion of the 3000, entered in the settings, so I could get on with testing another strategy, rather than waste time. If this wasn't an SQL server issue, then what was it?

Normally, I hope to be able to run GSB at my convenience, rather than when the SQL server is running smoothly. Hopefully, tomorrow will be a brighter day.

I will look into this. If you can get a screen shot of the 3 news stats from admin mode, would be helpful.

Offline Usage

JasonT - 16-9-2018 at 02:48 AM

Hi Peter, this is not a bug per se but more of a feature request.

When travelling and without access to an internet connection, when you attempt to load GSB, it can't complete its licence check, provides an error pop up and won't open. It seems you need an internet connection load GSB even if you don't need it to then operate and produce systems. This limits users' ability to continue to use GSB in those circumstances - which is a shame.

Can you please consider a method to enable offline usage, even if its for a limited time?

If an method or work around currently exists, can you please advise how as I will be in need of this capability over the coming month. An email or PM is fine if it is not appropriate to post it on the forum.

Many thanks,

Jason.

admin - 16-9-2018 at 04:43 PM

Quote: Originally posted by JasonT

Hi Peter, this is not a bug per se but more of a feature request.

When travelling and without access to an internet connection, when you attempt to load GSB, it can't complete its licence check, provides an error pop up and won't open. It seems you need an internet connection load GSB even if you don't need it to then operate and produce systems. This limits users' ability to continue to use GSB in those circumstances - which is a shame.

Can you please consider a method to enable offline usage, even if its for a limited time?

If an method or work around currently exists, can you please advise how as I will be in need of this capability over the coming month. An email or PM is fine if it is not appropriate to post it on the forum.

Many thanks,

Jason.

If you leave gsb open on your laptop, and put the laptop to sleep when done with it, this might fix or greatly reduce the problem. Many users might be envious of your problem, traveling around the world on holiday & playing with GSB. Hope you both have a great & very special holiday.

Bruce - 16-9-2018 at 05:37 PM

Hi Peter,

Running new simulations with Training/Test/Validation, 40/40/20 respectively. However, I'm not getting any validation results, either on the graph or the table below. What could be preventing that from being calculated? Thx.

admin - 16-9-2018 at 09:16 PM

Quote: Originally posted by TradingRails

Hi Peter,

Running new simulations with Training/Test/Validation, 40/40/20 respectively. However, I'm not getting any validation results, either on the graph or the table below. What could be preventing that from being calculated? Thx.

check fitness on the midle right optimzition field. If its blank then the probelm is drastic. No systems being produced. Go to a worker an right click the top diag field, then look what systems metrics are being produces. You filters might be too tight.
If stuck we can do team viewer.
You could have filter for 100 trades, and your not getting 100 trades in the 40% (esp if nth is not set to all)

Exception Errors

Bruce - 17-9-2018 at 12:02 AM

I appear to be getting a number of these alerts regularly, any idea what to do to address the exception?

Screen Shot 2018-09-17 at 5.56.42 PM.png - 127kB

admin - 17-9-2018 at 12:04 AM

Quote: Originally posted by TradingRails

I appear to be a number of these alerts regularly, any idea what to do to address the exception?

version 49.07 and before use massive ram for nth tests. fixed in 08.
the error is when I had to reboot the server after it crashed. Ran out of ram.
It wont happen again as I have no more 49.07 running.
Ignore the error as all systems that fail to upload get send as soon as GSB is online

Two graph empty in PA 20180913.1

Petzy - 17-9-2018 at 07:15 AM

In the report "Drawdown Graph" in Portfolio Analyst I get two graphs that are empty.
"Contribution ofEachSystemtoMaxDrawdowns" and "Contribution ofEachSystemtoAverage5WorstDrawdowns"

Attachment: Login to view the details

admin - 17-9-2018 at 03:58 PM

Quote: Originally posted by Petzy

In the report "Drawdown Graph" in Portfolio Analyst I get two graphs that are empty.
"Contribution ofEachSystemtoMaxDrawdowns" and "Contribution ofEachSystemtoAverage5WorstDrawdowns"

send me the files I dont have and I will look into it. make sure they are zipped up else in dropbox etc
Make sure you also copied ALL files from this build of PA, not just the exe file

Petzy - 18-9-2018 at 12:56 AM

I sent it via onedrive... thanks

admin - 18-9-2018 at 09:56 PM

Quote: Originally posted by Petzy

In the report "Drawdown Graph" in Portfolio Analyst I get two graphs that are empty.
"Contribution ofEachSystemtoMaxDrawdowns" and "Contribution ofEachSystemtoAverage5WorstDrawdowns"

The reason is the file name is too long. Fixed in build 20180919.1 (not released yet)
For now use shorter file name

boothy - 20-9-2018 at 04:11 PM

Hi Peter,

I'm not sure if this is a bug, but why is it that often the best fitness listed on the right hand GUI is much higher than what is in the unique systems?

admin - 20-9-2018 at 04:17 PM

Hi Peter,

I'm not sure if this is a bug, but why is it that often the best fitness listed on the right hand GUI is much higher than what is in the unique systems?

Best show an example where they dont match. Im rusty on on now, but sometimes the GUI in places can show the average of the 29,30,31 min bars which can lead to confusion as its not the same when you are looking at just the 30.
Has gold changed, in that last I looked everything was flat for the last year. The volatility was very low.

boothy - 21-9-2018 at 12:00 AM

Hi Peter,

I'm not sure if this is a bug, but why is it that often the best fitness listed on the right hand GUI is much higher than what is in the unique systems?

I have done another test. Just a simple test with 2 workers, ES.30 data1 and SPX.30 data2. No multiple time frames.

you will notice that the best fitneess in the centre GUI of manager is 17,006,804 but the best on the RH GUI is 21,533,203.

you will also notice if you add both worker unique systems = 3882 but on the manager says 3880. I have grace period set to 0.

Also attached is screen shot of the worker with fitness 21,533,203 and 1487 systems on RH GUI but only fitness 16,430,250 with 1487 systems in centre GUI.

Thanks.

admin - 21-9-2018 at 12:05 AM

Hi Peter,

I'm not sure if this is a bug, but why is it that often the best fitness listed on the right hand GUI is much higher than what is in the unique systems?

I dont see the issue. The manager has the higher fitness, and there is no screen shot of the other worker. The workers top fitness is in the worker.
2 systems will be dropped as they are duplicates.

boothy - 21-9-2018 at 12:59 AM

Why does the best fitness in the R/H GUI not match the best fitness in the centre GUI (unique systems) when sorted high to low.

Where is the system with fitness 21,533,203? its not in the list of unique systems in manager or in unique systems in the worker that created it.

boothy - 21-9-2018 at 01:02 AM

sorry here is attached screen shots

InkedFitness issue worker_LI.jpg - 2.2MB

admin - 21-9-2018 at 01:03 AM

Why does the best fitness in the R/H GUI not match the best fitness in the centre GUI (unique systems) when sorted high to low.

Where is the system with fitness 21,533,203? its not in the list of unique systems in manager or in unique systems in the worker that created it.

Thats a totally different matter.
A system may be created with fitness 21533203, but it fails to pass on other metrics. ie pearsons etc.
This field is very useful as a diagnostic tool. ie if its blank, gsb is not producing systems. If its positive but no systems, the filters are too tight or it needs more time.

boothy - 21-9-2018 at 01:04 AM

sorry here is attached screen shots

boothy - 21-9-2018 at 01:07 AM

oh ok, I always wondered why the best fitness often didn't show up.

sorry for multiple posts, my screen froze and didn't show it was posting.

Thanks!

boothy - 23-9-2018 at 07:19 PM

Hi Peter,

Ive been getting this exception notice this morning

Attachment: Login to view the details

admin - 23-9-2018 at 07:29 PM

Hi Peter,

Ive been getting this exception notice this morning

Forwarded to programmer. Im not sure what it is. Many exceptions are not critical.

boothy - 24-9-2018 at 07:34 PM

Hi Peter,

Ive been getting this exception notice this morning

Forwarded to programmer. Im not sure what it is. Many exceptions are not critical.

Hi Peter,

I think I may have figured it out, in my price data files I had data 1 and data 2 with different start dates, I have since tidied that up and all seems to be working ok.

Thanks.

admin - 24-9-2018 at 07:43 PM

Hi Peter,

Ive been getting this exception notice this morning

Forwarded to programmer. Im not sure what it is. Many exceptions are not critical.

Hi Peter,

I think I may have figured it out, in my price data files I had data 1 and data 2 with different start dates, I have since tidied that up and all seems to be working ok.

Thanks.

good to know thats fixed it, but GSB is designed to cope with this. I hadnt heard back from the programmer yet

cyrus68 - 3-10-2018 at 01:15 AM

I have datasets that end on September 2018. I generated systems on GSB with nth=1 and mode set to notrd. The majority of the systems stopped trading on 1 Sept 2015. A minority correctly generated trades up to the end of Sept 2018.

I checked the datasets, and there is no break or anomaly on 1 Sept 2015. It is impossible to do statistical tests with such results.

admin - 3-10-2018 at 01:31 AM

I have datasets that end on September 2018. I generated systems on GSB with nth=1 and mode set to notrd. The majority of the systems stopped trading on 1 Sept 2015. A minority correctly generated trades up to the end of Sept 2018.

I checked the datasets, and there is no break or anomaly on 1 Sept 2015. It is impossible to do statistical tests with such results.

what market?
happy to do TV session tomorrow. First think I would ask is what is the range after 2015.
ES had same issue 2005 2006
Regardless nth shouldnt depend on results after 2015

cyrus68 - 3-10-2018 at 01:46 AM

USO. Prices plunged in 2008 and again in 2014. The plunge in 2008 was particularly steep. In this run I included 2008. In a re-run I will exclude 2008.

The systems that generated results up to Sept 2018 don't look bad at all.

admin - 3-10-2018 at 01:48 AM

USO. Prices plunged in 2008 and again in 2014. The plunge in 2008 was particularly steep. In this run I included 2008. In a re-run I will exclude 2008.

The systems that generated results up to Sept 2018 don't look bad at all.

Your issue is what I would expect if you are training on a volatile period.

cyrus68 - 4-10-2018 at 06:32 AM

I am getting odd results when running walk forward. The graph looks normal. But, the details from row 6 downwards are either missing or wrong.

I have put the 30 min dataset in the "WF Price Data" field. The WF script is produced correctly for the 30 min. But the trade and performance reports are for WF 29 min.

The problems are the same for WF run on all systems.

admin - 4-10-2018 at 03:01 PM

I am getting odd results when running walk forward. The graph looks normal. But, the details from row 6 downwards are either missing or wrong.

I have put the 30 min dataset in the "WF Price Data" field. The WF script is produced correctly for the 30 min. But the trade and performance reports are for WF 29 min.

The problems are the same for WF run on all systems.

I havnt seen that before. Send me your teamviewer details and I will look at it.
There could be two issues. Try with live price data being set to 30 min.
That however wont fix the issue from run 6 onwards.
Wild guess is you have lots of years removed by the dates field, and GSB WF isnt aware of this.

cyrus68 - 5-10-2018 at 02:37 AM

I set 'Live Price data' and 'WF price data' both to 30 min. I get the same error.
Both were set before the run. As I understand it, you can change the 'WF price data' field after the run, but not the 'Live price data' field.
Dataset starts Jan 2009. Excludes 2008.

Don't know if your email works for TV.

admin - 5-10-2018 at 03:13 PM

I set 'Live Price data' and 'WF price data' both to 30 min. I get the same error.
Both were set before the run. As I understand it, you can change the 'WF price data' field after the run, but not the 'Live price data' field.
Dataset starts Jan 2009. Excludes 2008.

Don't know if your email works for TV.

Just email me your team viewer details when your on line again

cyrus68 - 15-10-2018 at 11:18 PM

I'm afraid the problem involving wrong walk forward calculations is a recurring one. I have used the default settings of OOS=20% and runs=10. The data ends 26 Sept 2018. As you can see, the data has been divided into 9 segments. The 10th run has the same data as the 9th. The calculation of the WF summary stats is likely distorted by the fake run.

Anybody running walk forward should check the details and not rely on the summary stats and the curves alone.

admin - 15-10-2018 at 11:21 PM

I'm afraid the problem involving wrong walk forward calculations is a recurring one. I have used the default settings of OOS=20% and runs=10. The data ends 26 Sept 2018. As you can see, the data has been divided into 9 segments. The 10th run has the same data as the 9th. The calculation of the WF summary stats is likely distorted by the fake run.

Anybody running walk forward should check the details and not rely on the summary stats and the curves alone.

my guess is no trades occurred in the last run. send me TV details and I will look.

admin - 20-12-2018 at 10:24 PM

50.45 and slightly older builds
I assume an issue with all builds that have favorites A & B
cant remove things from favorites
cant wf from favorites
stats didnt show count of how many entries until stats a and b were filled
Will get it fixed asap

engtraderfx - 21-12-2018 at 03:12 AM

Hi Peter, just trying to set up MC, ran script from 50.45 & copied code but got an error on "GSB_Decision2", wouldn't compile "unkown function".

Regards, Dave

admin - 21-12-2018 at 04:15 AM

Quote: Originally posted by engtraderfx

Hi Peter, just trying to set up MC, ran script from 50.45 & copied code but got an error on "GSB_Decision2", wouldn't compile "unkown function".

Regards, Dave

Did you import
C:\GSB\GSB (Managers)\Supplementary Scripts (TS & MC)\GSB_SCRIPTS_2018_12_15... .eld?

engtraderfx - 21-12-2018 at 04:57 PM

Thanks that did it, i just used the Supplement scripts from File menu option.

engtraderfx - 22-12-2018 at 07:12 AM

Getting this exception error tonight running on manager & cloud? Have killed GSB & rebooted for tonight

admin - 23-12-2018 at 03:37 PM

Quote: Originally posted by engtraderfx

Getting this exception error tonight running on manager & cloud? Have killed GSB & rebooted for tonight

That will mean an internet outage. GSB server was rebooted for 2 minutes a day or so ago too. If its not this then its on your end.
Let me know if its a problem still

engtraderfx - 25-12-2018 at 05:43 PM

thanks yes works now, I did have a lot of internet slowness that day, maybe the holiday cheer?!

admin - 28-12-2018 at 07:51 PM

50.54 installer file had a critical but simple bug of wrong path.
Manually change this path.
new installer soon

admin - 29-12-2018 at 06:17 AM

fixed in 50.55 installer

Trial has expired

awrch0n - 30-12-2018 at 02:33 AM

I want to try out the software. I have just downloaded the GSB 50.55 BETA BUILD Dec 29 2018 and upon running the application I get the "Trial has expired" message. Could you kindly provide a fix?

admin - 30-12-2018 at 03:36 PM

Quote: Originally posted by awrch0n

I want to try out the software. I have just downloaded the GSB 50.55 BETA BUILD Dec 29 2018 and upon running the application I get the "Trial has expired" message. Could you kindly provide a fix?

Please try to copy this file into c:\gsb and c:\gsb\cloud workers
If that doesnt fix it, email me your teamviwer.com details

Attachment: Login to view the details

awrch0n - 31-12-2018 at 01:43 AM

It is fixed now. Thank you.

"This application will not run from within a virtual machine"

awrch0n - 3-1-2019 at 09:49 AM

Hi Peter,

Build 50.55c

I'm getting the following error: "This application will not run from within a virtual machine".

My setup is a iMac running windows 10 on a Parallels Virtual Machine.

I am still on my trial period.

Could you please share with me the required licence file that resolves the issue?

admin - 3-1-2019 at 03:45 PM

Quote: Originally posted by awrch0n

Hi Peter,

Build 50.55c

I'm getting the following error: "This application will not run from within a virtual machine".

My setup is a iMac running windows 10 on a Parallels Virtual Machine.

I am still on my trial period.

Could you please share with me the required licence file that resolves the issue?

I will send you an email about this
Peter

admin - 9-1-2019 at 01:14 AM

There is a intermittent bug on 50.66 and possible many other builds.
A system from a random worker pops up in your manager. There may be no graph and metrics may change when you click on them.
This is the top priority to fix
Clue in screen shot here is user only build ng systems, not ES

Dates error

ProbTrader - 20-1-2019 at 04:22 PM

Hello

The dates functionality (IS/OOS) is still not working. The actual trades are also incorrect so the issue is not only graph related.
Please advise.

Thanks

admin - 20-1-2019 at 07:22 PM

Quote: Originally posted by ProbTrader

Hello

The dates functionality (IS/OOS) is still not working. The actual trades are also incorrect so the issue is not only graph related.
Please advise.

Thanks

This I hope will be fixed in 50.84. Its being worked on right now.
Programmer mentioned its due to some dates not being in either training or test period

coccigelus - 22-1-2019 at 10:19 AM

50.84.

I am getting a situation where i get same results under different metrics (net profit - DD - trades) during optimization through Manager and worker, even tough different indicators are employed;

Tradestation confirmed something wrong. I do not recall the same issue with 50.82

Carl - 22-1-2019 at 01:51 PM

Hi coccigelus,

I have seen this before. It is mathematically possible.
What other indicator are used in these three systems in your screenshot?

What results do you have in Tradestation if you use these three systems? Identical or different results?

coccigelus - 22-1-2019 at 02:30 PM

Quote: Originally posted by Carl

Hi coccigelus,

I have seen this before. It is mathematically possible.
What other indicator are used in these three systems in your screenshot?

What results do you have in Tradestation if you use these three systems? Identical or different results?

Hello Carl,

I am running right now version .82 and I am not seeing those discrepancies in the results. Ts has confirmed the issue. (different results between those systems)

admin - 22-1-2019 at 02:35 PM

coccigelus, You have 5 systems with the identical results. I think this is a bug. Sorting on fitness or pl will show how wide spread this is. Can you send me teamviewer.com details

coccigelus - 22-1-2019 at 02:54 PM

Peter, It was quite widespread . I have closed now .84 and using .82. Let me know If You want me to launch again and then we can keep in touch via skype - TW.

.82 seems working fine.

admin - 22-1-2019 at 03:01 PM

Peter, It was quite widespread . I have closed now .84 and using .82. Let me know If You want me to launch again and then we can keep in touch via skype - TW.

.82 seems working fine.

If your on the manager, check if you can see the same system in a worker.
The worker name should be in the code. If its one of your workers see if the results match. No other users have the issue, but I think its a bug.
If you see it on .82 let me know too. I will look via teamviewer.com

coccigelus - 27-1-2019 at 05:21 AM

Peter, It was quite widespread . I have closed now .84 and using .82. Let me know If You want me to launch again and then we can keep in touch via skype - TW.

.82 seems working fine.

Hello I am seeing the same issues with 51.01. Additional details: Although my worker is not very fast, I am not seeing those same results in the metrics. The issues come from workers that are not in my pc such as xeon 2690x2 etc.

I'll keep opened the manager and worker If You would like have a look via TW

admin - 27-1-2019 at 02:54 PM

Peter, It was quite widespread . I have closed now .84 and using .82. Let me know If You want me to launch again and then we can keep in touch via skype - TW.

.82 seems working fine.

Im happy to look at this with TV.

Random Workers

Bruce - 28-1-2019 at 07:31 PM

There is a intermittent bug on 50.66 and possible many other builds.
A system from a random worker pops up in your manager. There may be no graph and metrics may change when you click on them.
This is the top priority to fix
Clue in screenshot here is user only build ng systems, not ES

I've noticed this issue has cropped up again as per the attached screenshot. this time there is an associated system & graph.
I've performed a complete restart to see if this reoccurs in future tests.

Screen Shot 2019-01-29 at 5.56.04 AM.png - 863kB

admin - 28-1-2019 at 07:36 PM

Quote: Originally posted by TradingRails

Im keen to fix this bug, but its hard to replicate. Suspect its systems that are floating around after grace period has expired. Its good you can sort on the symbol and just delete them

coccigelus - 1-2-2019 at 11:04 AM

Thank You for bug fix. Very happy with 51.05

avatartrader - 3-2-2019 at 06:20 PM

Just a heads up about an issue in the contracts file that is currently distributed as the default - The value for "Ticks" for all EC (6E) and CD (6C) contracts should be 20000 and not 100000 since they trade in .00005 increments and not .00001 (when I built it out I was accidentally thinking of Forex).

Sorry about that-

admin - 3-2-2019 at 06:23 PM

Quote: Originally posted by avatartrader

Just a heads up about an issue in the contracts file that is currently distributed as the default - The value for "Ticks" for all EC (6E) and CD (6C) contracts should be 20000 and not 100000 since they trade in .00005 increments and not .00001 (when I built it out I was accidentally thinking of Forex).

Sorry about that-

Thanks for the tip. I will update GSB

avatartrader - 3-2-2019 at 06:39 PM

Not necessarily a correctable bug, but here is an issue that I experienced yesterday (which I was eventually able to resolve) that may have to do with how GSB managers/workers clear (or don't) cached or other in-memory system/price data between builds:

I was beginning to do some initial research and prototyping on the 6E to assess the viability of using some GSB to create some currency futures systems to add to my portfolio. As I just mentioned in the previous post, I discovered that I made an error in the contract specs, and so obviously, the first run of systems I created were invalid.

I stopped the build process, corrected the problem, updated the optimization settings, and started some new iterations. Even after the change, I was getting dramatically different results in TS/MC than GSB - even after restarting the manager.

Finally, I ended up clearing the price data cache on the manager, updating it with the most recent data, and starting a new iteration. This time, as I was clicking on the systems to review, many of the generated systems were recalculating as I was clicking on them, resulting in significant changes to the original stats as I went along.

However, this time both the recalculated systems and ones that didn't recalculate did match up when tested in TS/MC.

So, as best as I can tell, even though I had stopped the previous builds, there was something cached either on the manager or in the workers that was causing a significant number of systems to be generated and/or sent over that were generated based on bad data, probably associated with the original erroneous contract settings.

This was on 51.05 and all workers were being managed by RM. I didn't kill them between builds, so they definitely still had systems from prior builds.

At any rate, I am not having any issues now, but I thought I would mention it in case you think it is worth looking further into or anyone else has experienced similar behavior.

admin - 3-2-2019 at 07:13 PM

Quote: Originally posted by avatartrader

Not necessarily a correctable bug, but here is an issue that I experienced yesterday (which I was eventually able to resolve) that may have to do with how GSB managers/workers clear (or don't) cached or other in-memory system/price data between builds:

I was beginning to do some initial research and prototyping on the 6E to assess the viability of using some GSB to create some currency futures systems to add to my portfolio. As I just mentioned in the previous post, I discovered that I made an error in the contract specs, and so obviously, the first run of systems I created were invalid.

I stopped the build process, corrected the problem, updated the optimization settings, and started some new iterations. Even after the change, I was getting dramatically different results in TS/MC than GSB - even after restarting the manager.

Finally, I ended up clearing the price data cache on the manager, updating it with the most recent data, and starting a new iteration. This time, as I was clicking on the systems to review, many of the generated systems were recalculating as I was clicking on them, resulting in significant changes to the original stats as I went along.

However, this time both the recalculated systems and ones that didn't recalculate did match up when tested in TS/MC.

So, as best as I can tell, even though I had stopped the previous builds, there was something cached either on the manager or in the workers that was causing a significant number of systems to be generated and/or sent over that were generated based on bad data, probably associated with the original erroneous contract settings.

This was on 51.05 and all workers were being managed by RM. I didn't kill them between builds, so they definitely still had systems from prior builds.

At any rate, I am not having any issues now, but I thought I would mention it in case you think it is worth looking further into or anyone else has experienced similar behavior.

Lets keep an eye on this. Something to look at on the far right of the systems is the contract used. Check its the correct one, also not from another GSB user.
As far as I know this issue was fixed.
Also check if changes to nth / dates will make the stats = what they were before you clicked on them

engtraderfx - 18-4-2019 at 04:45 AM

Minor issue noted testing AD (Aus dollar futures), the trade list only shows results in 2 decimal places so can't see actual trade value, should be to at least same decimals as contract being used (4 or 5 for currency in this case). Chrs, Dave

admin - 18-4-2019 at 05:12 AM

Quote: Originally posted by engtraderfx

Minor issue noted testing AD (Aus dollar futures), the trade list only shows results in 2 decimal places so can't see actual trade value, should be to at least same decimals as contract being used (4 or 5 for currency in this case). Chrs, Dave

Agreed, thanks for that. Will fix before too long

appengineer - 24-5-2019 at 06:16 AM

Hi Peter,

I observed that when I run the wf-stats macro, the Global date and date range gets modified after the macros is complete

I started with End Global date of 2015-06-30 and date range of 1900-01-01 to 2015-06-30. This got updated to End Global date 2018-02-28 and range 1900-01-01 to 2018-02-28.

How does this change the result if I am rerunning wf-stats for Astab >=40, should I set it back to the original dates before rerunning macros or it doesn't impact the result?

Thank you

admin - 24-5-2019 at 06:19 AM

Quote: Originally posted by appengineer

Hi Peter,

I observed that when I run the wf-stats macro, the Global date and date range gets modified after the macros is complete

I started with End Global date of 2015-06-30 and date range of 1900-01-01 to 2015-06-30. This got updated to End Global date 2018-02-28 and range 1900-01-01 to 2018-02-28.

How does this change the result if I am rerunning wf-stats for Astab >=40, should I set it back to the original dates before rerunning macros or it doesn't impact the result?

Thank you

yes
global and the other dates back to 20150630
nth to all
oos off
wf param off.
I can send a macro tomorrow
off to sleep now

admin - 24-5-2019 at 05:17 PM

Quote: Originally posted by appengineer

Here are my macros.
Check them before use.
We now have neat mouse-overs that show us the content.
But we also have some sort of issue where numbers sneaked in. My guess is there might be a macro version compatibility issue

Attachment: Login to view the details

How to name price data file

appengineer - 26-5-2019 at 01:17 PM

Is this a real bug, see attached

My price data file is named as
NG1330.1.MINUTE.900-1430_CentralUsaTime_cl_ho_ng_rb_09-2006-to-date.txt

Carl - 26-5-2019 at 02:36 PM

I think you have to use Minute, not MINUTE

admin - 26-5-2019 at 05:46 PM

appenginer,
I think the issue is in the format of the data. Tools app settings data.
most MC/ts/quote.com formats are auto detect

appengineer - 27-5-2019 at 12:41 PM

appenginer,
I think the issue is in the format of the data. Tools app settings data.
most MC/ts/quote.com formats are auto detect

I exported the data from TS
Here is a sample
02/28/2008,14:24,18.553,18.553,18.532,18.532,29,91

And the settings
M/d/y,
yyyyMMdd,
d/m/y,
M/d/yyyy,
d/M/yyyy

H:m,
HHmm,
H:m:s,
HHmmss,
HHmmss fffffff

I will also try to change the file name from MINUTE to Minute when the build is done.

Thanks

admin - 27-5-2019 at 05:07 PM

appengineer
d/m/y is incorrect
its d/M/y //M is month., m is minute

see the last part of this in the docs
https://trademaid.info/gsbhelp/Data1.html
I asked the programmer, whats case sensitive in gsb.
Nothing except share keys, date time file formats under app settings , workplaceid, instantID, sharekeys
symbol.MINUTES.30 IS FINE IN UPPER and lower case

boothy - 29-5-2019 at 07:40 PM

Hi Peter,

I have just started up 52.45 to do some testing and GSB is stuck on status initializing and is not picking up any workers and is not starting to build systems. Its been initializing now for 15min.

admin - 29-5-2019 at 07:51 PM

Hi Peter,

I have just started up 52.45 to do some testing and GSB is stuck on status initializing and is not picking up any workers and is not starting to build systems. Its been initializing now for 15min.

Is that the wf status?
send me a screen shot and or teamviewer details

boothy - 29-5-2019 at 08:34 PM

Is that the wf status?
send me a screen shot and or teamviewer details

No it was status under workplace R/H GUI. It finally did start building systems but none had any system metrics.
I think its related to price files again. I updated my ES 1 min file but when I add it in and put multiplier of 30, as soon as I click OK it changes the time frame from 1 to 30.

GSB Worker - Guard page Error

appengineer - 30-5-2019 at 06:05 AM

Hi Peter,

See attached, is it because of running too many workers, running 10 workers on e5 2680, 256GB RAM

Thank you

admin - 30-5-2019 at 06:07 AM

Increase your virtual mem size and reboot.
Likely rm not configured correctly

admin - 30-5-2019 at 08:02 AM

Hi Boothy
that looks correct. your c/us figure is good. (lower is good)
1min * 30 is supposed to give 30
Im not sure if your stuck or resolved.

boothy - 30-5-2019 at 04:55 PM