| Pages:
1
..
8
9
10
11
12
..
17 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Quote: Originally posted by engtraderfx  | | I managed to get it down now to a few months by reloading data, eg CL data from IQFeed starts on 19/9/2006 with MC showing back to 18/12/2006 so
pretty close now. Have sent query to IQ on data order. I notice in my systems some have a large drawdown in that first 6 mths vs rest ( eg 3 to 4K
vs around 2K for rest of data), was wondering if its reasonable to ignore that early data. |
From memory most cl systems were bad in 2006. You can check on the demo CL system. The data was also a bit thin. So 18/12/2006 should be fine.
|
|
|
engtraderfx
Junior Member
Posts: 98
Registered: 15-10-2018
Member Is Offline
Mood: No Mood
|
|
Peter, I got a response from MC today with a hotfix that helped my reload data back to first data date. If anyone on MC wants the patch let me know.
For example,
ES 1m data back to 6/9/2005
CL 1m data back to 19/9/2006
Chrs, Dave
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by engtraderfx | Peter, I got a response from MC today with a hotfix that helped my reload data back to first data date. If anyone on MC wants the patch let me know.
For example,
ES 1m data back to 6/9/2005
CL 1m data back to 19/9/2006
Chrs, Dave
|
Hi Dave, thanks for the update
|
|
|
traderusa
Banned
Posts: 5
Registered: 24-5-2018
Member Is Offline
Mood: No Mood
|
|
GSB has some awesome features that no one else has. Things like being able to test 1000's of systems, to see objectively what works best.
The method to do this is fairly simple, but could it either have keyboard shortcuts, that programs like autoit could use to automate or could pre
configured commands be done in gsb.
for example
build 10,000 systems
when done, put into statsA
then change nth from do no trade to trade
Then put into stats B etc
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by traderusa | GSB has some awesome features that no one else has. Things like being able to test 1000's of systems, to see objectively what works best.
The method to do this is fairly simple, but could it either have keyboard shortcuts, that programs like autoit could use to automate or could pre
configured commands be done in gsb.
for example
build 10,000 systems
when done, put into statsA
then change nth from do no trade to trade
Then put into stats B etc
|
Excellent idea, and we can and will do it. Keep looking in the beta builds section details when its closer.
|
|
|
traderusa
Banned
Posts: 5
Registered: 24-5-2018
Member Is Offline
Mood: No Mood
|
|
Very impressive with the speed you got such a significant feature added - macro's. You have exceeded my expectations. 
|
|
|
coccigelus
Junior Member
Posts: 73
Registered: 11-7-2018
Member Is Offline
Mood: No Mood
|
|
Peter,
I am still exploring the custom indicator tool which I believe is a real important feature:
- Right now custom indicators is not currently supported for multi data optimization
- Validation for the reasons above is not possible.
The best solution would be have the feature to be able to import directly the function. But I understand that this is (probably) a lot of work.
However one thing that I think would be a very acceptable advanced features with little programming work required, would be import the value of the
custum indicator for each data stream with which we are interested to work. So let’s do an example:
We want to optimize ES30 SPX30 and we want to validate on Nq30 – NDX30.
Using the code generated by the software with little modifications we could run the custum indicators 4 times on each data stream set as data1 in
order to retrieve the value needed for each of the four data streams.
Once finished the process import the custum indicator that GSB should do automatically for each data stream.(checking each folder created into
Indicator data folder and checking the name of the custom function) So in this case, once GSB initialize, should raise a warning only when one
specific data stream does not have the value for the required custum indicator and/or data stream required, letting GSB users to have available all
the multidata stream and validation featureas available.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by coccigelus | Peter,
I am still exploring the custom indicator tool which I believe is a real important feature:
- Right now custom indicators is not currently supported for multi data optimization
- Validation for the reasons above is not possible.
The best solution would be have the feature to be able to import directly the function. But I understand that this is (probably) a lot of work.
However one thing that I think would be a very acceptable advanced features with little programming work required, would be import the value of the
custum indicator for each data stream with which we are interested to work. So let’s do an example:
We want to optimize ES30 SPX30 and we want to validate on Nq30 – NDX30.
Using the code generated by the software with little modifications we could run the custum indicators 4 times on each data stream set as data1 in
order to retrieve the value needed for each of the four data streams.
Once finished the process import the custum indicator that GSB should do automatically for each data stream.(checking each folder created into
Indicator data folder and checking the name of the custom function) So in this case, once GSB initialize, should raise a warning only when one
specific data stream does not have the value for the required custum indicator and/or data stream required, letting GSB users to have available all
the multidata stream and validation featureas available.
|
I will ask programmer about this.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi coccigelus
this can and will be done, but im not sure when. Trying to get the new secondary filters in GSB 2.0. So likely soon after this.
|
|
|
coccigelus
Junior Member
Posts: 73
Registered: 11-7-2018
Member Is Offline
Mood: No Mood
|
|
Excellent, Thank You!
|
|
|
Marc7200
Junior Member
Posts: 19
Registered: 9-11-2018
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
It would be nice if you would not have to create price data for every time frame.
Since you already have a contract list it should be possible to eliminate the price data list if the naming is the same as in the contract list.
If we only use 1 Minute price data files it would then be possible to simple change time frames 5min , 10min,.... GSB could do the work in the
background instead saving this information in a price data file?
Nice to have would be the option to test systems for example by saying 10-90min in increment of 5min and run all the possible time frames without
having to create so many price data files.
Also it would be helpful if there is one central location for price files not one for worker one for standalone and one for manager.
Just some ideas for improvement.
Thanks
Marc
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by Marc7200 | Hi Peter,
It would be nice if you would not have to create price data for every time frame.
Since you already have a contract list it should be possible to eliminate the price data list if the naming is the same as in the contract list.
If we only use 1 Minute price data files it would then be possible to simple change time frames 5min , 10min,.... GSB could do the work in the
background instead saving this information in a price data file?
Nice to have would be the option to test systems for example by saying 10-90min in increment of 5min and run all the possible time frames without
having to create so many price data files.
Also it would be helpful if there is one central location for price files not one for worker one for standalone and one for manager.
Just some ideas for improvement.
Thanks
Marc |
Hi Marc
Good to hear from you.
this can be done already
https://trademaid.info/gsbhelp/Data1.html
I probably should make this more obvious in the docs.
location of files.
With workers, you dont need to do anything.
Managers and standalone should be configured to use the same location. This should be the default config.
|
|
|
Marc7200
Junior Member
Posts: 19
Registered: 9-11-2018
Member Is Offline
Mood: No Mood
|
|
Hello Peter,
Is there by chance a video showing how this is done?
Thanks for the feedback and great work your doing!
Marc
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by Marc7200 | Hello Peter,
Is there by chance a video showing how this is done?
Thanks for the feedback and great work your doing!
Marc |
Thanks for your comments.
I will update the docs in a few days. Problem with videos is GSB 2.0 is in the pipeline, and this means many videos should be re done.
The main visual change I expect is new secondary filters(SF), optional but default of the existing SF to be 'locked' to each contract.
The logic in this is the number1 error GSB uses make is wrong SF.
Either the wrong closed, or all SF set to true.
All SF to be in the same list, including the non normalized
Maybe all SF to only be the closed variants, and then make the rest as tertiary filters
|
|
|
Marc7200
Junior Member
Posts: 19
Registered: 9-11-2018
Member Is Offline
Mood: No Mood
|
|
Wonderful looking forward to it!
|
|
|
waldocktrades
Junior Member
Posts: 14
Registered: 12-7-2018
Member Is Offline
Mood: No Mood
|
|
Peter, maybe this should be posted on the wish list.
It would be great to run a macro that culls the weak performers between Nth IS/OS, then double-blind date period, cull weak performers, run remainder
on more recent double-blind. Then, we could manually go through the survivors using data to current date.
I typically look at average trade. It would be REALLY nice to be able to compare individual system degradation, rather than a cutoff value. I
typically delete everything below $125 as too small of an average trade. I'll also delete systems below a given profit threshold based on the general
population. It would be really nice to know on a system by system basis and say only keep OS systems with an average trade within 10% of their IS avg
trd.
After all, the point of this pursuit is to find the most consistent algos over time.
52.06 is working fine. Good call.
Best, Andy
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by waldocktrades | Peter, maybe this should be posted on the wish list.
It would be great to run a macro that culls the weak performers between Nth IS/OS, then double-blind date period, cull weak performers, run remainder
on more recent double-blind. Then, we could manually go through the survivors using data to current date.
I typically look at average trade. It would be REALLY nice to be able to compare individual system degradation, rather than a cutoff value. I
typically delete everything below $125 as too small of an average trade. I'll also delete systems below a given profit threshold based on the general
population. It would be really nice to know on a system by system basis and say only keep OS systems with an average trade within 10% of their IS avg
trd.
After all, the point of this pursuit is to find the most consistent algos over time.
52.06 is working fine. Good call.
Best, Andy |
Hi Andy
I like what you propose. A number of users are saying some similar things, but from very different angles. This is similar to WF efficiency thats in
EWFO.
GSB macros does support metrics/favorites, but Ive never even looked or documented it. This feature in the macros was not intuitive and I want it
revamped.
We cant yet filter on fitness ration IS/OOS, but I want to do this.
Programmer says 52.06 has some significant bug in contract section, and all uses should be on 52.08. However I never noticed any bug.
Enjoy Easter.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Hi Andy
this is favorite in macro. I have not used it yet.
You can also remove from favorites too.
How I think it works is your criteria is added into favorite.
Then say you change dates and or nth,
then you run it again, and it will remove the systems that dont pass the metrics.
So if your first metric was pf >2,
you then change nth, and run pf >2
what remains is only systems that have pf2 with nth being true and not true.
I have never used this, but was told this by the programmer verbally.
|
|
|
waldocktrades
Junior Member
Posts: 14
Registered: 12-7-2018
Member Is Offline
Mood: No Mood
|
|
I do a lot of manual sorting based on the same ideas. Unfortunately, it forces us into thresholds, rather than being able to compare individual algo
IS/OS.
The current process works. My current algos represent the general populations, as they're designed to.
My belief is that being able to track the individual algo IS/OS will allow us to be able to move further up the performance ladder.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by waldocktrades | I do a lot of manual sorting based on the same ideas. Unfortunately, it forces us into thresholds, rather than being able to compare individual algo
IS/OS.
The current process works. My current algos represent the general populations, as they're designed to.
My belief is that being able to track the individual algo IS/OS will allow us to be able to move further up the performance ladder.
|
I like what you propose.
Short term, here is a possibility that you can modify to suit you.
I use 2000 to 2050630 nth 1, nth day 80.
Then turn all dates and nth to all. Then get the most linear systems using pearsons or pearsons date. You can then put a system into favoures, Run a
macro with stats to get the ratio. First change nth to no trade again, dates to trade, run stats a,c then nth to trade, statsB, dates to no trade,
nth all, stats d
Weakness is stats c,d are comparing years that are not equal.
Bottom line is we will code the feature in GSB, but need to get the first GSB 2.0 (at least) first
|
|
|
avatartrader
Junior Member
Posts: 60
Registered: 1-10-2018
Member Is Offline
Mood: No Mood
|
|
Just a small feature request:
It would helpful if there was an on-screen label or other indication of how many systems were selected in the Unique-Systems grid - for example, a
status bar docked to the bottom of the panel containing the tab control hosting the grid, or whatever is appropriate. Probably how I would do it is to
use a status bar control with a status label that only displays the selected total when the selected items are > 1 (e.g. 1000 of ).
When there is a very large number of systems, it is often hard to tell how many are being selected for walk forward, etc. and this would help when the
process prescribes "select approximately x systems", etc., particularly when you need to relax some of the criteria like NP/DD and so on to select an
ample amount for WF testing.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by avatartrader | Just a small feature request:
It would helpful if there was an on-screen label or other indication of how many systems were selected in the Unique-Systems grid - for example, a
status bar docked to the bottom of the panel containing the tab control hosting the grid, or whatever is appropriate. Probably how I would do it is to
use a status bar control with a status label that only displays the selected total when the selected items are > 1 (e.g. 1000 of ).
When there is a very large number of systems, it is often hard to tell how many are being selected for walk forward, etc. and this would help when the
process prescribes "select approximately x systems", etc., particularly when you need to relax some of the criteria like NP/DD and so on to select an
ample amount for WF testing. |
I agree. Excel has this. For now put them into favorites and it will give you the count. I will see if this can be done.
Macros I hope will do this in an up coming build.
|
|
|
JasonT
Junior Member
Posts: 61
Registered: 6-6-2018
Member Is Offline
Mood: No Mood
|
|
Data Visualisation
One of the challenges I have is understanding what the impact is of changes I make to settings in the Contracts List and Sessions List have on the
data that GSB uses to test systems against. For example changes to MOC, Last bar or Exch Time on the Contracts List and the open and close time on
the session list. I can imagine what they are supposed to do but I am not sure that what I am expecting is what is happening. As a result it reduces
the confidence I have in applying GSB generated systems to tradestation.
Is it possible to embed a chart visualisation tab to GSB to enable you to compare what data GSB is testing a system on to what you've got set up on
your trading software?
Perhaps it could be available using a Chart tab, next to the settings tab. Maybe when you click on the Trading Data or Opt Data that is the chart
data that is shown. Perhaps also (later) if you were to select a particular system that had been tested any indicators would be plotted and the
entries and exits could be plotted to enable even more effective debugging of the translation of systems between GSB and trading software.
Thanks for your consideration.
Jason.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by JasonT | One of the challenges I have is understanding what the impact is of changes I make to settings in the Contracts List and Sessions List have on the
data that GSB uses to test systems against. For example changes to MOC, Last bar or Exch Time on the Contracts List and the open and close time on
the session list. I can imagine what they are supposed to do but I am not sure that what I am expecting is what is happening. As a result it reduces
the confidence I have in applying GSB generated systems to tradestation.
Is it possible to embed a chart visualisation tab to GSB to enable you to compare what data GSB is testing a system on to what you've got set up on
your trading software?
Perhaps it could be available using a Chart tab, next to the settings tab. Maybe when you click on the Trading Data or Opt Data that is the chart
data that is shown. Perhaps also (later) if you were to select a particular system that had been tested any indicators would be plotted and the
entries and exits could be plotted to enable even more effective debugging of the translation of systems between GSB and trading software.
Thanks for your consideration.
Jason. |
You mean visualization of the charts, and indicators?
I need to think on this but I see some merit to it.
There is however extensive diagnostics built into gsb.
The main use is to compare ts to gsb. asci files written and can be compared in windiff
see this
files written in Data\Debugs
|
|
|
JasonT
Junior Member
Posts: 61
Registered: 6-6-2018
Member Is Offline
Mood: No Mood
|
|
Yes. Thanks for the consideration.
|
|
|
| Pages:
1
..
8
9
10
11
12
..
17 |
![[*]](images/davis/default_icon.gif){kind=icon}
