| Pages:
1
..
17
18
19 |
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@chuck,
I will look into this. I suspect its hard coded like you imply
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Hi Peter and Chuck,
I have noticed the same issue for GSB_SellRelativeStrength.
Only 20 bars can be used.
Thanks
Thanks received (1):
+1 admin at 2023-03-22 05:16:21
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
I just checked all 158 primary filters:
Same for:
GSB_BollingerBandHL (only 20)
GSB_ChaikinVolatility (only 9,9)
All the other ones are working okay.
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
Indicator issues
Thanks, Carl, for taking the time to do the research on indicators that appear to have hard-coded parameters. It sounds like you and I should/could
be working together. I was about half way through that job when your post arrived. I have been getting around the problem with a Tradestation
solution. But, it will be much better if Peter is able to get it fixed in GSB.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Exhaustive check done by the programmer
here is the total list that need to be fixed. Thanks for bringing this to my attention
BollingerBandsHL
ChaikinVolatility
PercentR
SellRelativeStrength
SmoothWaveIndex
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
For Chaikin's volatility, does anyone have an idea of the parmater ranges to be used for its two inputs?
I guess the average len is less than the roc??
CV = RateOfChange( GSB_Average( Range, MALength ), ROCLength ) ;
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
On Chaikin's website and in his webinars and in his book... this is what he recommends:
Chaikin’s Volatility is calculated by first calculating an exponential moving average of the difference between the daily high and low prices. Chaikin
recommends a 10-day moving average. Next, calculate the percent that this moving average has changed over a specified time period. Chaikin again
recommends 10 days.
To answer your question, however, I think a range of 2 to 20 should do nicely for both parameters.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
all this fixed in 65.49. We need longer lengths as most people are using intra day data.
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
Also since a couple of GSB version back the "force-use" setting does not work 100% correct anymore.
GSB Manager: only about 15% of the strategies contain the force-used indicator.
GSB Standalone: 100% correct, force-used PF is always present in the strategy
Thanks.
|
|
|
TwntySQ
Junior Member
Posts: 26
Registered: 17-2-2023
Member Is Offline
|
|
Hi!
Tried WF yesterday with som systems and same problem with the WF stopping at 93-94%.
Any fix for this available?
|
|
|
TwntySQ
Junior Member
Posts: 26
Registered: 17-2-2023
Member Is Offline
|
|
Hi!
Tried WF yesterday with som systems and same problem with the WF stopping at 93-94%.
Any fix for this available?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@carl, nothing has changed in the section for years
you cant force more than 3 indicators
@twntySQ
this is fixed in the next release which is under testing
Ive been slow to release it as old mangers will not talk to new workers due to compatibility issues.
Still doing some tweaks to vwap code too.
There is a bug cover up for the issue. Under custom indicators remove (12 or 13) from the length of the offending indicator.
I forgot what indicator had this issue.
(cant remember if its 12 or 13 to be removed. worst case remove both
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
I know, I used 3 indicators with only one indicator force-use.
I haven't used "force-use" recently. The last GSB results I could find on my desktop when "force-use" results were correct, were from June 2022. GSB
version 64.97.
Could it be it is not a GSB issue, but a SQL server issue?
Because force-use works fine with GSB Standalone, but not with GSB Manager.
Thanks
|
|
|
Carl
Member
Posts: 342
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Hi Peter,
The "force-use" option really has a bug.
I ran several tests and the bug is there all the time when using GSB Manager.
Thanks
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@carl, thanks for your observation. This should make is easier to diagnose
|
|
|
TwntySQ
Junior Member
Posts: 26
Registered: 17-2-2023
Member Is Offline
|
|
Hi ,
Have tested the dirty fix for WF not completing. Also reduced gen and population as shown in earlier posts.
After iWF finishes it says "CanceledLost" in the WF cell.
Any fixes for that?
Thanks in advance.
|
|
|
TwntySQ
Junior Member
Posts: 26
Registered: 17-2-2023
Member Is Offline
|
|
Hi ,
Have tested the dirty fix for WF not completing. Also reduced gen and population as shown in earlier posts.
After iWF finishes it says "CanceledLost" in the WF cell.
Any fixes for that?
Thanks in advance.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@twntySQ
Cloud WF lost/deleted from DataBase.. Likely worker killed/closed or worker didn't send updates for long (because on internet failure or else) and
DataBase did auto cleanup
thats a small risk of using public cloud for wf
Not sure if resubit works
Otherwise save system, load system in new manager and wf.
you can wf to local manager or try the cloud again
Thanks received (1):
+1 TwntySQ at 2023-04-04 08:04:45
|
|
|
TwntySQ
Junior Member
Posts: 26
Registered: 17-2-2023
Member Is Offline
|
|
Hi again,
Thanks for the answer Peter.
Tried to load the systems in manager and then in a standalone version but after removing the custom indicator as mentioned earlier, the WF goes to 92%
and then I get the following message as attached.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@TwntySQ
there are work rounds for this.
you either remove custom parmater lenght of 10 or 11 (forgot what one, but you can do both) on the indicators that are failing wf
or I get you the pre-release build that has the fix.
THere is no cloud power for this build yet (apart from your own), but likely I release it in a week or so.
|
|
|
REMO755
Member
Posts: 181
Registered: 11-4-2021
Member Is Offline
|
|
Hello,
I have problems in Tradestation with the version.
GSB_SCRIPTS_2021_11_11+WITH UPDATEDGSBSYS1EN_V1.21
Value1 = GSB_Scripts_2021_11_11;
I send a photo with the errors.
what is the problem?
Please help.
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
issues with parameter ranges
Firstly, let me say that this is not a NEW issue with the latest release. It's been there forever.
The automatic generation tool available in "custom indicators" is a magical way to generate parameter ranges for one, two or even three parameters.
The problem is that there is a limit of 200 combinations. If you want to vary parameter one from 1 to 50 by 1 and parameter two from 1 to 4 by 0.5,
that results in 650 combinations and the tool won't accept more than 200.
So, you have to compromise... incrementing the first value by 2 and the second value by 0.5, for instance.
Interestingly, the text window to the left of the generation tool can handle at least 1,000 combinations and they all work perfectly. I use it every
day. So far, I haven't had to the need to go beyond 1,000 so I don't know the real limit.
Since the text window will handle at least 1,000 combinations, can we change the limitation on the generate tool from 200 to 1,000?
For what it is worth, I generate my total combinations 200 at a time, copy them to notepad, appending as I go. I then copy the entire set from
notepad to the text window. Works perfectly, but it would be SO much easier if that 200 limit was increased. Especially since the underlying logic
is capable of handling a lot more.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@chuck
the 200 limit is very deliberate limit. Too many combinations is too hard work for GA
Not a great work around, but you could build a set of 200 variables, see how well it works.
Then make a different 200 limit and compare.
Not saved systems will not match if you change these variables to something different.
I will enquire of the programmer next time I chat over the possibility of expansion.
You can also optimize a system in ts using wider parameters. Using coarse values in optimization
if quite ok often. Fine values give increase in sample results and decreased OOS results.
|
|
|
ChuckNZ
Junior Member
Posts: 50
Registered: 22-12-2022
Member Is Offline
|
|
More on parameter selection
Hi Peter and thanks for your comments. I agree with you 100% and this will seem counter intuitive to many.
When I ultimately generate strategies, I want fewer rather than more possible parameters and that prompts me to ask for the limit of 200 to be
increased. Seems strange, right? The default range for many of the indicators is 1 to 200 incrementing by 1.05%.
I go through a process for every timeframe for every instrument that I want to trade.
First, I want to know "what are the best settings" for each indicator? To do that, I will ran a process where I turn off all filters and stops and
let gsb try every possible setting. That will mean at least 200 possible strategies. If there is an ATR multiplier or some other parameter, the
200 could easily become 800. Also you apply offsets to some of the indicators.
I do a strategy creation run for just one indicator which doesn't take very long to try every possible value or set of values. I then sort by my
favourite fitness function and delete the bottom 2/3 of the results.
I then prepare a bell curve for the parameters that are in the 1/3 that performed well. I remove the outliers and bing, bang, boom I have a subset of
values that I will use when generating strategies.
Instead of gsb wasting time on settings between 1 and 200, I might be using values between (say) 45 and 70.
This process has improved my results considerably according to every metric.
Now you have my "secret sauce". I hope you and others find it to at least be interesting.
Quote: Originally posted by admin  | @chuck
the 200 limit is very deliberate limit. Too many combinations is too hard work for GA
Not a great work around, but you could build a set of 200 variables, see how well it works.
Then make a different 200 limit and compare.
Not saved systems will not match if you change these variables to something different.
I will enquire of the programmer next time I chat over the possibility of expansion.
You can also optimize a system in ts using wider parameters. Using coarse values in optimization
if quite ok often. Fine values give increase in sample results and decreased OOS results. |
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@chuck, the idea is interesting, & time consuming. I like James clears book Atomic habits
Big increases in performance can be the result of cumulative small improvements.
As long as your idea was only done with the in sample data, it looks to have merit. (that critera is critical to me)
Semi related, I have also considered having multiple of one indicator
ie
rsi5-50
rsi40-100
rsi80-200
Might that be a really fast way of doing the same thing?
Thanks received (1):
+1 TwntySQ at 2023-04-20 12:39:59
|
|
|
| Pages:
1
..
17
18
19 |
![[*]](images/davis/default_icon.gif){kind=icon}
