| Pages:
1
..
21
22
23
24
25
..
98 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Quote: Originally posted by appengineer  | Peter,
After additional testing and trial, looks like "Show Systems on Workers GUI" property has to be set to True. If I set it to False no systems are
generated.
Thank you |
That is supposed to hide systems on the worker, not the manager. I will look into that. Might just remove the option.
|
|
|
fra2019
Junior Member
Posts: 9
Registered: 28-4-2019
Member Is Offline
Mood: No Mood
|
|
Anyone got the following error message when trying to automate a GSB system in TradeStation?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
your code has > 1 data stream, but chart has only one
ie you might have $spx for data2 etc
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
Peter,
GSB Manager became unresponsive after generating about 48000 systems (max set 50000). I looked in my resource monitor and looks like it's still doing
something but after hours it's still unresponsive, can I kill the process.
Will it be able to resume where it left?
Thank you
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by appengineer | Peter,
GSB Manager became unresponsive after generating about 48000 systems (max set 50000). I looked in my resource monitor and looks like it's still doing
something but after hours it's still unresponsive, can I kill the process.
Will it be able to resume where it left?
Thank you |
What version of GSB are you on?
There was a fix for high ram and cpu a few versions ago.
Have you ran out of ram? How much ram do you have?
There is an option to save systems if certain performance figures are reached, but I have it turned off myself as long term it would use a lot of my
dropbox space
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi appengineer
RM is more focused on worker cpu, though it shows total cpu. Best check the ram and cpu useage in windows task manager
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
Peter,
My computer ran out of RAM (90% in use), it has about 8GB, GSB Manager is using about 6GB.
I managed to pause the process at about 49,100 systems built.
I will let it run overnight.
I am unable to connect to the server to monitor the workers but I think GSB is processing, as it is writing to disk.
Thank you
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by admin | Hi appengineer
RM is more focused on worker cpu, though it shows total cpu. Best check the ram and cpu useage in windows task manager |
GSB Manager
RAM is 90%+ used, CPU 5% and disk write/read 100%
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by appengineer | Peter,
My computer ran out of RAM (90% in use), it has about 8GB, GSB Manager is using about 6GB.
I managed to pause the process at about 49,100 systems built.
I will let it run overnight.
I am unable to connect to the server to monitor the workers but I think GSB is processing, as it is writing to disk.
Thank you
50k systems on 8GB ram is not a good combination.
You can reduce the amount of cores used under app settings, which might use less ram at the cost of speed.
Would be good if you can upgrade the ram or machine
|
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by appengineer | Peter,
My computer ran out of RAM (90% in use), it has about 8GB, GSB Manager is using about 6GB.
I managed to pause the process at about 49,100 systems built.
I will let it run overnight.
I am unable to connect to the server to monitor the workers but I think GSB is processing, as it is writing to disk.
Thank you
50k systems on 8GB ram is not a good combination.
You can reduce the amount of cores used under app settings, which might use less ram at the cost of speed.
Would be good if you can upgrade the ram or machine
|
|
Peter,
Can GSB leverage GPU?
Thank you
|
|
|
fra2019
Junior Member
Posts: 9
Registered: 28-4-2019
Member Is Offline
Mood: No Mood
|
|
When I download historical data from TradeStation, time are based on my computer time i.e. market opens at 21:30 and closes at 04:00 the following
day. I understand I need to change the contract table with MOC From 0400 to 0459. Can GSB understand this is following day (so for instance, a session
would be Monday 21:30 to Tuesday 04:00) ?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by fra2019 | | When I download historical data from TradeStation, time are based on my computer time i.e. market opens at 21:30 and closes at 04:00 the following
day. I understand I need to change the contract table with MOC From 0400 to 0459. Can GSB understand this is following day (so for instance, a session
would be Monday 21:30 to Tuesday 04:00) ? |
I think this can be done.
There are similar issues for me in Australia, so I keeping my trading computer in central usa time. This fixes other issues like daylight savings. usa
and au change at different times. There are also rare cases where closed(1) doesnt work properly in australia time zone
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
Looking at "Proving the methodology" article, we are using
0% of the data for Test(Optimization settings)
but the Performance Filter for Test is set to
Min Pearson:0.9,
Min Profit Factor:1.5,
Min Trade Count: 100)
Are these settings correct? I was under the impression setting Test % to 0, will result to no test phase but 100% training with nth out of sample.
Also when enabling the best indicators for ES, I don't see any indicators prefixed with GSB_ do I need to load them as custom?
Or by selecting AverageFc that equates to GSB_AverageFc
Thank you
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by appengineer | Hi Peter,
Looking at "Proving the methodology" article, we are using
0% of the data for Test(Optimization settings)
but the Performance Filter for Test is set to
Min Pearson:0.9,
Min Profit Factor:1.5,
Min Trade Count: 100)
Are these settings correct? I was under the impression setting Test % to 0, will result to no test phase but 100% training with nth out of sample.
Also when enabling the best indicators for ES, I don't see any indicators prefixed with GSB_ do I need to load them as custom?
Or by selecting AverageFc that equates to GSB_AverageFc
Thank you |
All settings used are here
https://trademaid.info/gsbhelp/Provingthemethodology.html
Set pearsons to 0.95, and min trade count to 150
Test is ok at 0, as we are using nth and dates to get OOS periods.
There is still a number of things I need to refine and test further.
This is what I am testing to see if it helps.
Using only + operatrs and weights of only 1.
SO no "-" operand, and no weights -1,1,step2
Note also that when you get the top 2000 persons systems, and the take the top 1000 np/dd of the 2000, there is very little over lap between choosing
np/dd first vs pearsons.
I dont yet know what is best if any
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by appengineer | Hi Peter,
Looking at "Proving the methodology" article, we are using
0% of the data for Test(Optimization settings)
but the Performance Filter for Test is set to
Min Pearson:0.9,
Min Profit Factor:1.5,
Min Trade Count: 100)
Are these settings correct? I was under the impression setting Test % to 0, will result to no test phase but 100% training with nth out of sample.
Also when enabling the best indicators for ES, I don't see any indicators prefixed with GSB_ do I need to load them as custom?
Or by selecting AverageFc that equates to GSB_AverageFc
Thank you |
All settings used are here
https://trademaid.info/gsbhelp/Provingthemethodology.html
Set pearsons to 0.95, and min trade count to 150
Test is ok at 0, as we are using nth and dates to get OOS periods.
There is still a number of things I need to refine and test further.
This is what I am testing to see if it helps.
Using only + operatrs and weights of only 1.
SO no "-" operand, and no weights -1,1,step2
Note also that when you get the top 2000 persons systems, and the take the top 1000 np/dd of the 2000, there is very little over lap between choosing
np/dd first vs pearsons.
I dont yet know what is best if any
|
Thank you Peter!
Is the idea to follow the same principle(proving the methodology) and apply it to different markets, as we tweak the settings and see what works best?
Thank you
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Appengineer, absolutely correct. So this also lets you validate what I have done. I have done 5000 WF in the last day, so lots more to investigate.
So far it looks like + operand with no "-" operand and weights all 1 are best.
There is a lot more scope to tweak. IE do we wf with weights 1, or a range.
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
Proving the methodology
| Quote: Originally posted by admin | Appengineer, absolutely correct. So this also lets you validate what I have done. I have done 5000 WF in the last day, so lots more to investigate.
So far it looks like + operand with no "-" operand and weights all 1 are best.
There is a lot more scope to tweak. IE do we wf with weights 1, or a range. |
I built about 50,000 ES systems using the settings described in the proving the methodology document but I am getting different results. see attached.
I have observed that the GSB process has stopped but the Manager is using a lot of RAM. The Last Update timestamp keep on updating making me believe
there is some process still going on.
Any idea why the degradation result differs significantly?
By the way I used the 16 indicators listed on proving the methodology
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Appenginer
there is something very wrong. The macro should give stats a to f, which you dont have.
Stats G is something I just added. It is the stats of all 50,000 systems with nth all, dates 2000 to 20150630
If you compare stats A to StatsG, you can see how the 5000 systems I picked compared to the 50,000
(mildly interesting but not important at this stage.
Best you send me teamviewer details.
I suspect you are not running the same macro as me.
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by admin | Appenginer
there is something very wrong. The macro should give stats a to f, which you dont have.
Stats G is something I just added. It is the stats of all 50,000 systems with nth all, dates 2000 to 20150630
If you compare stats A to StatsG, you can see how the 5000 systems I picked compared to the 50,000
(mildly interesting but not important at this stage.
Best you send me teamviewer details.
I suspect you are not running the same macro as me. |
I think you are right, I didn't use the correct macro and my global end date was also off (2018-02-28 rather than 2015-06-30)
Let me give it one more try if it is still different I will send u a teamviewer
|
|
|
appengineer
Junior Member
Posts: 61
Registered: 8-4-2019
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
How can you tell when GSB has finished building systems.
I did set the Max number of unique systems to 50,000.
The status says Stopped, but the number of unique systems still keeps on going up - currently at 53,940 unique system built
Thank you,
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by appengineer | Hi Peter,
How can you tell when GSB has finished building systems.
I did set the Max number of unique systems to 50,000.
The status says Stopped, but the number of unique systems still keeps on going up - currently at 53,940 unique system built
Thank you, |
set grace period under app settings to say 60, = 60 seconds.
this ignores all systems coming in after its stopped.
However grace period doesn't work perfectly.
It doesn't cause problems with the stats though, but I put a delay in the macro to over come this small bug.
A warning, don't cancel macros. If you do, no future macros run. Will get this fixed asap
|
|
|
avatartrader
Junior Member
Posts: 60
Registered: 1-10-2018
Member Is Offline
Mood: No Mood
|
|
This is not so much a support issue as a general question:
Once I have completed a build process, walked forward systems, applied the settings and filtered the systems with astab >= 40, I like to save the
systems for later review and analysis, or for future selection for monitoring/trading.
When I save the systems, it appears that the current applied parameters are serialized out with the system correctly (using current WF parameters and
OOS = true), but it does not seem to serialize the setting as to whether or not the current WF parameters are applied or a flag that it was walked
forward at all.
I realize this is mainly cosmetic as long as it saves and re-applies the right parameters when I re-open the system. Right now, my process is to only
save the top walked forward systems anyway, so I just wanted to make sure I am understanding correctly and that what I am saving for later review and
use are maintaining all of the prior work that I did to build and walk them forward.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by avatartrader | This is not so much a support issue as a general question:
Once I have completed a build process, walked forward systems, applied the settings and filtered the systems with astab >= 40, I like to save the
systems for later review and analysis, or for future selection for monitoring/trading.
When I save the systems, it appears that the current applied parameters are serialized out with the system correctly (using current WF parameters and
OOS = true), but it does not seem to serialize the setting as to whether or not the current WF parameters are applied or a flag that it was walked
forward at all.
I realize this is mainly cosmetic as long as it saves and re-applies the right parameters when I re-open the system. Right now, my process is to only
save the top walked forward systems anyway, so I just wanted to make sure I am understanding correctly and that what I am saving for later review and
use are maintaining all of the prior work that I did to build and walk them forward. |
The most important thing is you save the eld, and you also want to save the WF version of the eld. My practice is to save it into TS. I didnt think
the WF parameters save, but I havnt checked. The GSB system can be re-created out of the binary hash at the end of the code. But its nice still to
save the GSB system too. Does this answer your question?
|
|
|
avatartrader
Junior Member
Posts: 60
Registered: 1-10-2018
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by admin | | Quote: Originally posted by avatartrader | This is not so much a support issue as a general question:
Once I have completed a build process, walked forward systems, applied the settings and filtered the systems with astab >= 40, I like to save the
systems for later review and analysis, or for future selection for monitoring/trading.
When I save the systems, it appears that the current applied parameters are serialized out with the system correctly (using current WF parameters and
OOS = true), but it does not seem to serialize the setting as to whether or not the current WF parameters are applied or a flag that it was walked
forward at all.
I realize this is mainly cosmetic as long as it saves and re-applies the right parameters when I re-open the system. Right now, my process is to only
save the top walked forward systems anyway, so I just wanted to make sure I am understanding correctly and that what I am saving for later review and
use are maintaining all of the prior work that I did to build and walk them forward. |
The most important thing is you save the eld, and you also want to save the WF version of the eld. My practice is to save it into TS. I didnt think
the WF parameters save, but I havnt checked. The GSB system can be re-created out of the binary hash at the end of the code. But its nice still to
save the GSB system too. Does this answer your question? |
Yes it does. I didn't save the EL code this last time, but I do have a couple of the systems that I had previously ported the EL code for, so I can
compare and see what it actually saved since those are using the WF.
What is extremely strange is that neither of the id's of the systems I exported as EL match any of the systems I saved, even though I know for a fact
I saved off the same systems.
Even the date portion of the id is wrong. the saved systems all have 20190507, which is when they were built, but the EL code for the couple that I am
testing is 20190508.
So, I think you must be right - it looks like the WF EL code is given it's own unique ID different from the system ID, and that is not preserved for
sure. Unfortunately, since the ID of the original system and the WF code I do have is different, I can't easily go back and find the system to see if
the current parameters after loading are what were the WF parameters at the time.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Avatartrader,
I think the ID is made with a date/time stamp in it to avoid duplicate system names. Best save your top systems into ts and or save them with a
specific name in GSB.
ie my topESsystem1
|
|
|
| Pages:
1
..
21
22
23
24
25
..
98 |
![[*]](images/davis/default_icon.gif){kind=icon}
