| Pages:
1
..
28
29
30
31
32
..
98 |
Sten
Junior Member
Posts: 35
Registered: 25-10-2019
Member Is Offline
|
|
Quote: Originally posted by LucaRicatti  |
Hi Stan,
I just endend my trial period and decided to purchase GSB and apply for a Tradestation account (got Multicharts but they need to fix data2 bug).
I had poor results with trial version and had difficulty understanding the method like you, but I liked so much the robustness-stress idea behind GSB
that I choose not to give up and asked Peter for help.
Here the results of my - assisted by Peter - last build with trial version:
PS: First time in this business i met someone dedicated like Peter.
|
Hi, LucaRicatti!
Thank you very much for your feedback!
TradeStation does not work with the residents of the country where I live, so I have to use MultiCharts (and something like IQFeed for historical
data).
Out of curiosity, what is MultiCharts data2 bug you are referring to? I'm a long time MultiCharts user, and I'm not aware of any data2 bugs. I've
heard of TradeStation's data2 related bug, on the contrary.
P.S. Yes, Peter is very responsive and provides a lot of support.
|
|
|
LucaRicatti
Junior Member
Posts: 16
Registered: 12-10-2019
Member Is Offline
|
|
Sten check the last post of Beta Builds topic for MC data2.
I’m on Interactive Broker + MC
|
|
|
LucaRicatti
Junior Member
Posts: 16
Registered: 12-10-2019
Member Is Offline
|
|
edit - double post
|
|
|
Sten
Junior Member
Posts: 35
Registered: 25-10-2019
Member Is Offline
|
|
| Quote: Originally posted by LucaRicatti |
I had poor results with trial version and had difficulty understanding the method like you, but I liked so much the robustness-stress idea behind GSB
that I choose not to give up and asked Peter for help.
Here the results of my - assisted by Peter - last build with trial version:
|
Hi LucaRicatti,
Can you share your optimization settings from your trial version? I want to compare them to mine.
|
|
|
LucaRicatti
Junior Member
Posts: 16
Registered: 12-10-2019
Member Is Offline
|
|
| Quote: Originally posted by Sten |
Hi LucaRicatti,
Can you share your optimization settings from your trial version? I want to compare them to mine.
|
Hi Sten,
I've opened the settings you attached:
- Dates and Global Dates: both ends at 2015-6-30
- In WF Params: all WF Indic (1,2,3) Weight Min is 0 (not 0.5)
- Max Uniq Systems: 50.000
- Too much Indicators, i built with my best 9
|
|
|
Sten
Junior Member
Posts: 35
Registered: 25-10-2019
Member Is Offline
|
|
LucaRicatti, thank you! I'll try these settings.
I did not formally verify which indicators are the best, but looked at the table in documentation, and selected all that were marked as green (PF >
1). Probably, I should run my own tests and select the best indicators.
|
|
|
Sten
Junior Member
Posts: 35
Registered: 25-10-2019
Member Is Offline
|
|
Peter,
Here is another severe bug in GSB.
GSB Manager 1.0.55.90 / 2019-11.07
Windows 10 x64
I started GSB Manager and loaded previously saved 248 strategies. I then add all strategies to Fav.A, sort them on PAS column and add all strategies
with PAS > 40 to Fav.B. Then I load and run wf_stats macro for Fav.A. While macro is running I get an exception message among other messages. Here is
an exception:
Attachment: Login to view the details
After macro finishes GSB goes into a "broken" state. When I select a different strategy in Unique-Systems window, most of the time GSB does not update
strategy code and PnL Graph:
This can easily be seen when I switch to Scripts tab and look at the strategy ID in the script code. Sometimes GSB switches strategy (but very
rarely). And sometimes I get a new exception while selecting a different strategy.
Probably this is the similar bug to the one I reported before, but now I know the exact steps how to reproduce it.
Attachment: Login to view the details
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by Sten | Peter,
Here is another severe bug in GSB.
GSB Manager 1.0.55.90 / 2019-11.07
Windows 10 x64
I started GSB Manager and loaded previously saved 248 strategies. I then add all strategies to Fav.A, sort them on PAS column and add all strategies
with PAS > 40 to Fav.B. Then I load and run wf_stats macro for Fav.A. While macro is running I get an exception message among other messages. Here is
an exception:
After macro finishes GSB goes into a "broken" state. When I select a different strategy in Unique-Systems window, most of the time GSB does not update
strategy code and PnL Graph:
This can easily be seen when I switch to Scripts tab and look at the strategy ID in the script code. Sometimes GSB switches strategy (but very
rarely). And sometimes I get a new exception while selecting a different strategy.
Probably this is the similar bug to the one I reported before, but now I know the exact steps how to reproduce it.
|
Hi Sten,
Is this bug repeatable? Do not do other things in gsb while macros are running - if you've had a hang in the past. The reason is a thread lock might
occur if you change something, and the macro is doing a similar change.
IU will look into the exception. Many exceptions can be ignored.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
hi Sten,
we need a support ticket with a system that gives this fault. Someone once said, sloppy language leads to sloppy thinking. (I forgot who it was)
A severe bug is where GSB doesnt work at all, all ts <> gsb (miles out). A intermittent macro hang is very annoying, but its not a severe bug.
Regardless we are committed to bug fixes 
I run that macro many times in a day, and dont get hangs. Thats not saying you dont get hangs, but its not a common bug thats easy for me to
reproduce.
|
|
|
Sten
Junior Member
Posts: 35
Registered: 25-10-2019
Member Is Offline
|
|
| Quote: Originally posted by admin | hi Sten,
we need a support ticket with a system that gives this fault. Someone once said, sloppy language leads to sloppy thinking. (I forgot who it was)
A severe bug is where GSB doesnt work at all, all ts <> gsb (miles out). A intermittent macro hang is very annoying, but its not a severe bug.
Regardless we are committed to bug fixes
I run that macro many times in a day, and dont get hangs. Thats not saying you dont get hangs, but its not a common bug thats easy for me to
reproduce. |
Yes, the bug is easily reproducible. I notice it very often, and was able to reproduce it from the first attempt using the steps I outlined above.
I respectfully disagree. After this bug occurs, it is impossible to continue to work in GSB Manager as it does not display a reliable results (does
not show correct system PnL chart, system code e.t.c.). Thus a bug is a severe one.
You probably need to use the same build as me to repeat this bug.
I attached all 248 systems that I initially load into GSB. As well as optimization settings that I use. wf_stats macro that I run can be found in my
post above.
Attachment: Login to view the details
Attachment: Login to view the details
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by Sten | | Quote: Originally posted by admin | hi Sten,
we need a support ticket with a system that gives this fault. Someone once said, sloppy language leads to sloppy thinking. (I forgot who it was)
A severe bug is where GSB doesnt work at all, all ts <> gsb (miles out). A intermittent macro hang is very annoying, but its not a severe bug.
Regardless we are committed to bug fixes
I run that macro many times in a day, and dont get hangs. Thats not saying you dont get hangs, but its not a common bug thats easy for me to
reproduce. |
Yes, the bug is easily reproducible. I notice it very often, and was able to reproduce it from the first attempt using the steps I outlined above.
I respectfully disagree. After this bug occurs, it is impossible to continue to work in GSB Manager as it does not display a reliable results (does
not show correct system PnL chart, system code e.t.c.). Thus a bug is a severe one.
You probably need to use the same build as me to repeat this bug.
I attached all 248 systems that I initially load into GSB. As well as optimization settings that I use. wf_stats macro that I run can be found in my
post above.
|
what build are you on? I may supply you with more recent build as we work hard to get rid of bugs, and so some problems disappear in new builds
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Sten, your settings are very close to what I use. (good), your weight 3 is wrong. Its blank and should be the same as weight1&2
On wf too. This is a minor point and wont make a big difference, or cause your bugs. Human error in settings is something ive done too often, and im
resolved now to be more methodical to avoid some of these traps. Ive wasted a lot of time this week over those sorts of things. windiff is good to
compare settings that you think should work, but don't with those that you know did work.
Here is an example of my own mistakes I picked up. These comments are for all users, as it may be helpful for others.
I want to do a video on windiff at some stage. I also use to diagnose ts <> (ts/mc)
GSB has extensive diag's to identify this sort of thing
|
|
|
Sten
Junior Member
Posts: 35
Registered: 25-10-2019
Member Is Offline
|
|
Hi Peter,
Thanks for the tip on weight 3. I'll fix it when I do next optimization.
Out of curiosity, I decided to research why strategy that I imported from GSB into MultiCharts does not generate any trades. I found this particular
strategy contained these two lines of code:
| Code: |
// Indicators
v1 = GSB_PriceLessPrevAverageFc(Volume, i1length) of Data(i1Data);
v2 = GSB_PriceOverPrevAverageFc(Volume, i2length) of Data(i2Data);
....
|
The strategy is built on 30 min bars, and is using Volume. The problem is that in MultiCharts Volume is not defined for intraday time frames, and we
need to use Ticks keyword instead.
After I rewrote the strategy like this it started to produce some trades:
| Code: |
var: intrabarpersist myVolume(0);
if BarType >= 2 and BarType < 5 then
myVolume = Volume
else
myVolume = Ticks;
// Indicators
v1 = GSB_PriceLessPrevAverageFc(myVolume, i1length) of Data(i1Data);
v2 = GSB_PriceOverPrevAverageFc(myVolume, i2length) of Data(i2Data);
|
You may want to fix MultiCharts code generation in GSB to correctly handle volume.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
HI Sten,
I dont use MC, but TS has similar structure. This needs to be fixed in TS and MC then. Thanks for spotting it. TS has two methods of volume, under its
chart properties. The volume functions performed poorly in indicator testing (surprisingly to me), but this might explain why.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Sten, I was also thinking - good to reduce the chance of human error in GSB. I see no reason why all weights are not made the same as the default
option.
all wf weights also the same, but they could be different (wider) than the build weights.
Ive made this mistake too often, and other gsb users likely did, though may not be aware of it.
Your bug is fixed in 56.10. Under testing now
|
|
|
emsjoflo
Junior Member
Posts: 39
Registered: 7-7-2017
Member Is Offline
Mood: No Mood
|
|
I was in the middle of a test with GSBmanager and when RM updated the workers, they no longer recognize my manager. I was wondering why my test was
taking so long and I see that I only have one cloud worker under my manager. I tried adding the gsbcloud3_password1234,
gsbcloud1_password1234 to my workers but then they get picked up by the cloud. Then i tried changing the group to GSBuser but that didn't work
either. any tips? I'd rather not abort the test I have in progress
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by emsjoflo | I was in the middle of a test with GSBmanager and when RM updated the workers, they no longer recognize my manager. I was wondering why my test was
taking so long and I see that I only have one cloud worker under my manager. I tried adding the gsbcloud3_password1234,
gsbcloud1_password1234 to my workers but then they get picked up by the cloud. Then i tried changing the group to GSBuser but that didn't work
either. any tips? I'd rather not abort the test I have in progress |
WHAT VERSION IS YOUR MANAGER, AND WHAT VERSION ARE THE WORKERS.
NORMALLY OLD MANAGERS TALK TO NEW WORKERS, AND NEW MANAGER WONT TALK TO OLD WORKERS.
|
|
|
emsjoflo
Junior Member
Posts: 39
Registered: 7-7-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by admin |
WHAT VERSION IS YOUR MANAGER, AND WHAT VERSION ARE THE WORKERS.
NORMALLY OLD MANAGERS TALK TO NEW WORKERS, AND NEW MANAGER WONT TALK TO OLD WORKERS. |
The manager was version 1.0.55.89. I aborted the test and started a new one with the latest version of the manager and it is working now
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
| Quote: Originally posted by admin | HI Sten,
I dont use MC, but TS has similar structure. This needs to be fixed in TS and MC then. Thanks for spotting it. TS has two methods of volume, under its
chart properties. The volume functions performed poorly in indicator testing (surprisingly to me), but this might explain why.
|
Peter, how should we interpret this? that volume data supplied from user in pricedata file has not been able to be utilised at all for MC users?
Thanks
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by Daniel UK1 | | Quote: Originally posted by admin | HI Sten,
I dont use MC, but TS has similar structure. This needs to be fixed in TS and MC then. Thanks for spotting it. TS has two methods of volume, under its
chart properties. The volume functions performed poorly in indicator testing (surprisingly to me), but this might explain why.
|
Peter, how should we interpret this? that volume data supplied from user in pricedata file has not been able to be utilised at all for MC users?
Thanks |
GSB will corectly use the volume in the way the data was originally exported.
However its using the EL word volume, which means up volume. I think it should be using upvol+downvolume which is the el word ticks. Im going to get
this changed before to long. This doesnt matter righ tnow as indicators with volume are not in the top 9 or so. (see indicator testing methodology)
By the way, the next video is now very close. 95% or more of the content is done, and it just needs refining / editing. Hope its out next week some
time.
It will be roughly 30 min long. Found one tweak to significantly help OOS results too.
|
|
|
JasonT
Junior Member
Posts: 61
Registered: 6-6-2018
Member Is Offline
Mood: No Mood
|
|
Self Help Actions when GSB produces no systems
Hi Peter,
I've just downloaded the latest version of GSB 56.010, and using the out of the box default settings and data, and am not able to get GSB to produce
any systems. I've reset app and optimization settings to default and tried it after restarting the computer.
Can you please provide a list of things we can check ourselves when this occurs?
Many thanks.
Jason
|
|
|
JasonT
Junior Member
Posts: 61
Registered: 6-6-2018
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by JasonT | Hi Peter,
I've just downloaded the latest version of GSB 56.010, and using the out of the box default settings and data, and am not able to get GSB to produce
any systems. I've reset app and optimization settings to default and tried it after restarting the computer.
Can you please provide a list of things we can check ourselves when this occurs?
Many thanks.
Jason |
So I've tried a few things to get it to work. One of the troubleshooting tests I tried was to adjust the Session that I had applied to my data under
Tools > Price Data. Initially that did nothing. Then all of a sudden after a change to one of the session options GSB started pounding out new
systems. Then I tried all sessions I had previously tried (0830-1500, 24/7, MOC @ 23:59) and all of them worked. Very weird.
Of note was that when I applied a session of 0830-1500 to my 0830-1500 CT data, it builds systems much faster than if I apply a 24/7 session to my
0830-1500 data. I'd be interested to know if anyone else has experienced the same.
|
|
|
emsjoflo
Junior Member
Posts: 39
Registered: 7-7-2017
Member Is Offline
Mood: No Mood
|
|
Is there a way to force use of secondary data?
Is there a way to force use of secondary data in creating systems? If not, is there a way to sort the systems by whether Data2 is being used or not?
I want to test the quality of systems that use both Data1 and Data2.
If not, I suppose I could possibly convert Data2 to a custom indicator and try to force use (in all systems) but the last time I tried that, GSB
ignored the forced use option and made systems with and without my "forced" custom indicator.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
my reply in CAPS
YOUR IDEA IS WORTH DOING AND SIMPLE TO DO.
| Quote: Originally posted by emsjoflo | Is there a way to force use of secondary data in creating systems? If not, is there a way to sort the systems by whether Data2 is being used or not?
NO
I want to test the quality of systems that use both Data1 and Data2.
EASY TO DO. FOLLOW THE METHADOLGY AND GET YOUR TOP 250 OF 50,000 SYSTEMS. THEN SAVE INTO EXCEL & LOOK AT THE % OF SYSTEMS IN YOUR TOP 250 THAT HAVE
DATA2
If not, I suppose I could possibly convert Data2 to a custom indicator and try to force use (in all systems) but the last time I tried that, GSB
ignored the forced use option and made systems with and without my "forced" custom indicator. |
NOT SURE WHY THAT WOULD HAPPEN. YOUR FIRST OPTION IS THE SIMPLE ONE
|
|
|
JasonT
Junior Member
Posts: 61
Registered: 6-6-2018
Member Is Offline
Mood: No Mood
|
|
Which systems go into stats?
Hi Peter,
If you build say 50,000 systems then (either manually or with a macro) put statistics into StatsA, which 10,000 systems get saved?
eg first 10,000 built? best 10,000 according to profit factor? best 10,000 according to fitness? etc
Thanks,
Jason
|
|
|
| Pages:
1
..
28
29
30
31
32
..
98 |
![[*]](images/davis/default_icon.gif){kind=icon}
