GSB Forums

GSB BETA BUILDS

 Pages:  1  ..  10    12    14

Daniel UK1 - 12-4-2023 at 07:27 AM

Hi Peter, I just saw that Nick confimed after testing multiple indicators still not matching, I propose those are fixed before pushing out this version, so we are sure that this version actually has the found miss matches fixed. No matching indicators will do no one good. Cheers


admin - 12-4-2023 at 05:01 PM

@daniel,
I am aware of this,but the new build is significantly better than the last build, and it has lots of other bug fixes that affect many users.

ie wf do not complete, false errors about sf must be enabled etc.
Im hoping to get more fixed before release
Indicator match issues have taken up months of programmer time, and we are still not finished.

You are still welcome to use the old build.

admin - 12-4-2023 at 08:39 PM

@daniel,
next to be fixed is
GSB_AccumDistDR
GSB_LowestLowDLessClose

Then maybe these two next:
GSB_PriceLessPrevAverageFc
GSB_PriceOverPrevAverageFc

I am trying to work on ones that have a lot of errors..

admin - 12-4-2023 at 10:04 PM

GSB_AccumDistDR is now a match in the next build to be released

admin - 13-4-2023 at 01:04 AM

There will be a new build released to all sunday night usa time.
It wont work with older builds.
Here is why
if we fix a indicator, ie GSB_AccumDistDR
and the old worker, gives system with GSB_AccumDistDR in it, but results are wrong and dont match fixed newer build of GSB
then we have a miss-match due to version compatibility.



So after this sundays build, I will not likely push out new build until we have 100% match on all indicators

Daniel UK1 - 13-4-2023 at 03:04 AM

Quote: Originally posted by admin  
There will be a new build released to all sunday night usa time.
It wont work with older builds.
Here is why
if we fix a indicator, ie GSB_AccumDistDR
and the old worker, gives system with GSB_AccumDistDR in it, but results are wrong and dont match fixed newer build of GSB
then we have a miss-match due to version compatibility.



So after this sundays build, I will not likely push out new build until we have 100% match on all indicators


Sounds good Peter, Thank you

admin - 14-4-2023 at 05:41 PM

GSB_PriceLessPrevAverageFc and GSB_PriceOverPrevAverageFc are now matching, to be released in the next build

admin - 16-4-2023 at 05:02 PM

65.56 released via auto updates to all beta testers.
vip to all.
you should have your managers closed, gsb automation not running.
If they are open, files cant be over written on auto update


new eld to be loaded
C:\GSB\Supplementary Scripts (TS & MC)\GSB_ALL_FUNCTIONS_2023_04_15_1.eld

admin - 16-4-2023 at 07:01 PM

65.56 released to all purchased users, but not trial users.

admin - 22-4-2023 at 12:29 AM

Next build
Genetic day of week and time of day Filter seems to be working, but might want to make this work with zero indicators purely to test the time of day / day of week bias

LowestLowDFc function is now a perfect match GSB TO TS

portfolioquanttrader2020 - 23-4-2023 at 11:01 AM

hello peter
What exactly does that filter do? is it secondary? Or is it multiplied with indicators?
Thank you

admin - 23-4-2023 at 08:17 PM

Next build
perfect match on
HighestHighDFc
HighestHighDLessClose
CloseLessHighestHighD
LowestLowDFc
LowestLowDLessClose
CloseLessLowestLowD
CloseToHHDLLD

admin - 23-4-2023 at 08:22 PM

@portfolioquanttrader2020
time of day is simply
if x >0
if time is startofsessiontime+x bars then .... buy

if x <0
if time is startofsessiontime-x bars then .... sell



same thing with dayofweek

if y >0
if dayofweek=y then .... buy

if y <0
if dayofweek=-y then .... sell
these are genetically chosen
Purpose is for gsb to find time of day / day of week bias in a market







admin - 7-5-2023 at 05:20 PM

next build under testing
perfect match fix on
GSB/Indicators/CloseHDLDminusClosev2.cs
GSB/Indicators/CloseHDLDminusClosev3
GSB/Indicators/CloseLessHighestHighD
GSB/Indicators/CloseLessLowestLowD
GSB/Indicators/CloseToHHDLLD
GSB/Indicators/HighestHighDFc
GSB/Indicators/HighestHighDLessClose
GSB/Indicators/LowestLowDFc
GSB/Indicators/LowestLowDLessClose

admin - 7-5-2023 at 09:39 PM

correction, GSB_HighestHighDLessClosev3 does not match yet

admin - 10-5-2023 at 12:31 AM

Next build, balanced the following GSB to TradeStation:
S2R2
S3R3
S2R2LessClose
S3R3LessClose
I wont release in a hurry, as all users will need to update to use the same version of manager to worker
Problems generally rounding issues, or wrong values from ts on the first bar

admin - 11-5-2023 at 12:06 AM

65.70 (see above ) has gone to beta testers.
Next we are working on catchup of some of the recent matchfix / ts code changes / updates to Ninja Trader

admin - 16-5-2023 at 04:31 PM

65.71 under testing shortly
match fix for 12 indicators
I will not release quickly as all users must upgrade to keep compatibility with the cloud
LessCloseS1R1
LessCloseS2R2
LessCloseS3R3
R1
R2
R3
S1
S1R1
S1R1LessClose
S2
S3

admin - 17-5-2023 at 10:35 PM

to be tested 65.74 / 2023-05-18: Custom Filter Parameters range is now editable
Spending a lot of time experimenting with various filters to test in GSB

admin - 19-5-2023 at 01:11 AM

Out of 194 indicators (I have a few not released etc) there are 4 bad matches in the next build, 3 with minor issues and a number with tiny rounding issues (no big deal)

Been a interesting but slow week.
If learnt a lot about filters, and learning - is this a curve fit, or a valid filter, and looking at for indices.
also have learn how to make some significant architectural changes to secondary filters and other filters.
Starting to experiment with put some of this into GSB shortly

indicator-match.png - 97kB

admin - 26-5-2023 at 05:23 PM

not yet tested
65.80
fix to trailing stop logic
filters now can have custom lengths
debug output for filters added
GSB_CloseDLessHighDLowDNormFilter to allow long and short to be possible (long or short set by secondary filter and primary indicators)
minor bug fixes

I will not release this build quickly, as all users will have to upgrade to use the cloud with the most recent version

admin - 26-5-2023 at 05:30 PM

@daniel, no idea why but you need to get the link going to beta builds and subscribe to that.
this has all the details.
4 inidcators match so bad must never be used. 30 or so with tiny rounding errors, 3 with mild errors but still useable

missmatch2.png - 405kB

Daniel UK1 - 28-5-2023 at 05:14 AM

Quote: Originally posted by admin  
@daniel, no idea why but you need to get the link going to beta builds and subscribe to that.
this has all the details.
4 inidcators match so bad must never be used. 30 or so with tiny rounding errors, 3 with mild errors but still useable


@Peter, many thanks for your kind reply.

I spoke to @Nick and those results from you, must be from a version that only you have, and not an official release. What Nick tested was the official latest pushed out release. The errors was much much larger than 4.

Is there perhaps a misunderstand things from my side if so, kindly let me know what version you are referring to that you showing results for ?

Perhaps @Nick can chip in that is running the tests

NickW - 28-5-2023 at 07:54 AM

I am on version 65.56 and use this version to exclude any indicators i found not working 100%

Daniel UK1 - 30-5-2023 at 07:59 AM

Thanks, Yes i think then that @Peter must be referring to a non official beta release, that is yet to be pushed out. Because the official pushed out 65.56 does not have only 4 failing indicators, that's for sure.

@Peter It would be great, if the indicator would have same name in stats export and it has inside GSB gui, as of now, its a bit of guessing when one tries to turn off and on indicators based on stats for failing non matching indicators.

admin - 30-5-2023 at 05:23 PM

I tested the latest release, but indicator match list is regularly published.
My match checks will differ as my getdiff.exe counts 2.0000000 the same as 1.99999999 etc, while a text compare wont get this tolerance correct

I agree the fact that some names dont match the eld name is annoying.
I will try to find a fix for this or a table lookup

admin - 2-6-2023 at 07:51 PM

under testing 65.81
Perfect match now on new filter, and new diag features for filters.
Release will still not be quick as all users need to upgrade to be compatible with the newer versions

admin - 12-6-2023 at 08:11 PM

65.86 under testing
Fix for systems that change when you click on them. This has been a bug for a very long time and it was very hard to fix.
New filter closelesshighdlowd matchs now too,
more internal diagnostics
Lack of updates for some time was due to a lot of time going into this issue.

admin - 27-6-2023 at 10:08 PM

Next build has export of nth statistics, and added export of filters info added
Next build of automation has adjustable nth mode, but the excel macro sheet has not yet got this feature added.
Im not releasing the build in a hurry as every user will have to upgrade to keep cloud compatibility

filters.png - 63kB

admin - 10-7-2023 at 08:19 PM

next build
export of mutiple trade lists
Export of wf trade lists
numerous bug fixes & small improvements.
NEW eld with fixed or updated indicators
Still not releasing as entire cloud needs to update to be compatible.
Likely a few weeks away

New settings for Gold


wf-tradelist.png - 81kB

admin - 11-7-2023 at 12:41 AM

also Save WF Trade List in macro ExportTrades
Import indicator default path not the same as export indicator default path

admin - 18-7-2023 at 01:15 AM

65.93 under testing but will not be next release build
exported into stats
stop and pt value
how often operands are used. ie if you have say multiply and divide
remove any . in eld names to keep compatibility with Ninja Trader. IE close.neg would now be closeneg
new eld's will be provided here/ These can be used for all builds of GSB for now.



newgsb.png - 63kB

Attachment: Login to view the details


admin - 21-7-2023 at 02:21 AM

New in 66.02 - but not being released yet
We made significant GSB progress this week, and after a month of efforts got digital file signing working again. (keeps anti virus programs happier)


Corelation.png - 94kBCorelation2.png - 139kB

GSB 66.02

thowoc213 - 21-7-2023 at 02:29 AM

Hi Peter,
Correlation functionality, that's good!
Will this work also on the saved systems?
When do you think you will release new version?
Best regards,
Thomas

admin - 21-7-2023 at 02:33 AM

@thowoc213
Release should be < 1 month. Im waiting for Ninja trader improvements, and more indicator matching.
Another feature / warning.
On a few occasions i have had a manager export with 0 systems in fav ABCD and the oos period F. Likely its due to me canceling the wrong manager by mistake.
However when you have a result of zero, this screws your stats if your trying to see if you last tweak was good or not.
Default on this setting is NOT USED, but I recommend putting a 1 here. - so no output in CSV if no systems

exports.png - 187kB

admin - 31-7-2023 at 10:06 PM

66.11 under testing. If anyone wants a copy, let me know
Has long short graph + metrics in GUI for long short
minor bug fixes
all vwap indicators now match. (there was only an issue at the very beginning of the data)

long.png - 263kB

admin - 4-8-2023 at 07:12 PM

I hope to release the next major release in 1 to 2 weeks. No complaints from those who have trialed the current build which is a good sign.
Whats important to note is when I do the upgrade, your old manager will NOT talk to new workers, so if your part way through a build this might bite you.
One solution is get the build from me before I release it via auto update

Im getting regular GSB updates of new tweaks and features, but keen to push out the next build

Im still spending much of my time on new methodology tweaks. Extremely CPU intensive and I dont have the the users donated cloud power as the old build will not work with new workers

admin - 8-8-2023 at 06:54 PM

this is the error missmtach of the new release over 4 years of data.
3 indicators have been removed
If it never got better than this, then this is exceptable but we will still seek to improve on this


Attachment: Login to view the details


admin - 11-8-2023 at 12:58 AM

66.22 under testing
It has close to perfect match on all indicators, and some new indicators under testing
Excel file of results included. I am still trying to improve on this but if it never got better, its very acceptable.
Anyone who wants this build, just email me
Also has days of week enabled exported into stats.
I will write more on that on the methodology thread

Highlights of the last month
I have repeatability within 1.7% using 192 indicators, which I am just delighted about.

Im also going to release some more indicators, but not sure what ones yet.


Attachment: Login to view the details


admin - 11-8-2023 at 03:10 AM

current eld are here, but there will be updates regularly short term

Attachment: Login to view the details


admin - 15-8-2023 at 11:32 PM

Im going to release build 66.25 to all but trial users in < 24 hours.
Note that if you have a manager running, and the workers update, the old manger will not talk to new workers (i might see if I can fix this)
new eld are included. This is a major update. If anyone wants the new build earlier, email me

admin - 16-8-2023 at 06:06 PM

66.25 just released via auto update
I can forsee some issues
there is new error checking and some earlier config files might show up as not valid
ie
emd 1 minute times 30 is valid
emd 5 min time 6 is not valid. An error will show when you run GSB

whenever a bar time is multiplied, the original bar must be 1 minute

admin - 17-8-2023 at 06:43 PM

next build
66.26
improved gsb to ts match on the following
new eld will be required
ChaikinOsc
DMI
DmiPlus
DmiMinus
Adx
AdxR




engtraderfx - 18-8-2023 at 07:58 PM

re error "whenever a bar time is multiplied, the original bar must be 1 minute" can confirm i am getting this on every datastream not 1-min. All my config files & saved systems are using M30 data so affecting every run now.....

admin - 18-8-2023 at 08:19 PM

Quote: Originally posted by engtraderfx  
re error "whenever a bar time is multiplied, the original bar must be 1 minute" can confirm i am getting this on every datastream not 1-min. All my config files & saved systems are using M30 data so affecting every run now.....

this should only affect you if you have m30 x2 for example.
this has to me removed /. changed if its the case

look out for emd data under verify. This is wrong in the actual installer file from a long time ago

if stuck we can do anydesk.com session

WFA stability

thowoc213 - 19-8-2023 at 03:59 AM

Peter,
any chance you could reveal the formula for stability calculation?

We are recommended to use systems with high stability.
At the same time it's hard to understand how the stability figures have arrived.

Below is an example of the results from a WFA run.
There are 4 parameters optimised and the stability results are 80 and 54.

Opt. 111 43 0 0 0 0 0 112 -1 47.5
1 142 24 0 0 0 0 0 88   -1 42.5
2 96   20 0 0 0 0 0 88 -1 65  
3 96   20 0 0 0 0 0 88 -1 65  
4 181 20 0 0 0 0 0 88   -1 65  
5 181 20 0 0 0 0 0 88   -1 65  
6 135 20 0 0 0 0 0 88   -1 62.5
7 135 20 0 0 0 0 0 88   -1 62.5
8 53   20 0 0 0 0 0 118 -1 40  
9 59   20 0 0 0 0 0 118 -1 47.5
10 53   20 0 0 0 0 0 118 -1 22.5
Curr 135 24 0 0 0 0 0 118 -1 22.5

Params. Rol. Stability Coarse: 80
Params. Anc. Stability Coarse: 54





admin - 21-8-2023 at 01:39 AM

@thowoc213
read this. Note we now user astab / rstab coarse which ignores insignificant changes. ie a length of 99 changing 100 are considered to be stable as its a 1% change.
a change from 9 to 10 is not stable as its a 11% move

This question should ideally be under https://trademaid.info/forum/viewthread.php?tid=229 as its not about beta build


stabilty.png - 254kB

admin - 21-8-2023 at 01:40 AM

Next release of GSB match check going from worst to best. If it never gets better than this, results are highly acceptable.

matchcheck.png - 101kB

admin - 23-8-2023 at 01:45 AM

this is the next release match check of ts to gsb worse at the top. Includes some inidcators that were disabled in the past as the match was too bad


matchcheck2.png - 70kB

admin - 10-9-2023 at 06:17 PM

66.38 under testing.
near perfect match on all indicators. Match check work is now finished unless new issues reported.

Includes error trap for when no data exists in the session time used.

matchechk.png - 39kB

Systemholic - 11-9-2023 at 02:28 AM

hi Peter latest ver 35 script file does not have this function code. Any chance to upload here for the benefit of everyone-

GSB_CloseOverPrevCloseDv2


Thanks

admin - 11-9-2023 at 03:44 AM

@systemholic
see here https://trademaid.info/forum/viewthread.php?tid=229&page=78
Likely I will release nte next build tommrow with it included

admin - 15-9-2023 at 12:43 AM

build 66.400 sent to beta testers
has new features for next build of GSB automation
Has latest ELD GSB_ALL_FUNCTIONS_2023_08_30_3 but no need to update unless you get a compile error in TS editor
All indicators under testing have 1 or less errors in TS match to GSB
Minor fixes

admin - 22-9-2023 at 01:01 AM

66.40 released. This is final release to get 99.9% match of all GSB to TS indicators

admin - 5-10-2023 at 11:05 PM

next build, one new indicator, 5 others were tried, but they were not good.
Improvements in Ninja Trader features that are in TS but not NT.
bug fixes

highlow10.png - 218kB

admin - 13-10-2023 at 01:11 AM

Next build fixes for changes in TS code, that wont compile in MC
Additional NT secondary filters and miss match fixes.
Hope to test in some time next week.

admin - 20-10-2023 at 05:47 PM

Working on ATR stop, break even stop, and another filter.
Next build has minor tweaks and NT improvements.

admin - 3-11-2023 at 05:49 PM

atr stop has taken a great deal of work to get right, but its close to release
A great deal of work has gone into long only 23 hours a day 5.5 days a week nq system.
These are going well, and many close or on new highs


https://trademaid.info/gsbhelp/BuildingNasdaq23hrX55daySwing...
https://trademaid.info/forum/viewthread.php?tid=117&page=38

Ive also tested and see that 60 minute bars work fine. faster but less liquidity, 20 min bars slower but better liqudiy and even more memory hungry.

Unrelated
the new intel Intel Core i9-14900 is out
a little faster than the 13900 and the 14900 takes 192 gb or ram, while 13900 takes 128 gb
This will be the fasted GSB machine, but the dual 2699v4 / 2696v4 intel (second hand) will be best bang for buck.

As the new i9 cpu gets maxed out with so few workers, you likely dont need >128 gb of ram, but always good if you can plan to expand should you need it.



admin - 13-11-2023 at 10:28 PM

66.56 release
this has new beak even profit lock, variable points locked in profit lock, and atr stop.
Know issue that exits that are stopped have there exit name called break even stop, when its the full stop loss, not BE stop that hit
That will be fixed in next build
new eld is in c:\gsb folder and only for some users is it in c:\gsb\Supplementary Scripts (TS & MC)\
new eld also are here

Break even stop seems to make significant difference in the amount of systems that appear in fav BCD in the new long only nq 23 hour 5.5 day systems.
Doing more testing...



Attachment: Login to view the details


Daniel UK1 - 15-11-2023 at 04:51 AM

@Peter, seems like you are doing lots of great improvement, thanks for that... May i ask .. the new stops, are they also tested to be working in live executions, since many exit types that existed in GSB did not really work in live trading, and for example, you many times ended up with one hard stop and another vol stop .. etc.. So these ATR and Breakeven stops, are they tested also to be fully working in live execution and not placing double stops when in one position etc.. ?

admin - 15-11-2023 at 05:29 PM

@daniel
sorry for you, but this logic has not changed. I asked programmer what would be involved to centralize stop logic - and the reply was I need to think on it.
Note in current build there is a bug where all stop exits are labeled profitlock if profitlock and fixed stop is used. This is fixed in the next release, but the current logic would suit you as its one stop signal instead of 2.
Just a tiny change in the code did this. However no idea how all the other stops interact. It took so long to get profit locks, and atr stop as there was a lot of complexities involved.
While I know this is important to you, there are work arounds to this issue on MC and there are much bigger issues I need to work on that affect a lot more users. Apologies.

Daniel UK1 - 16-11-2023 at 07:02 AM

@Peter Many thanks for your feedback.
Ask @Randy for the exact fix for resolving the not wanted two stops in place, perhaps he can help your dev.

I am not sure I understand your correctly here, " It took so long to get profit locks, and atr stop as there was a lot of complexities involved."
Does this mean that profit lock and atr stop, is now working in terms of logic in real trading?

admin - 16-11-2023 at 11:45 PM

I did more testing.
All the stop types work, but GSB does only use a max of one of these at a time.
atr stop, atr trailing, profit lock, breakeven stop.
(You can select them all, but GA will only choose one. I want to change this so multiple can be done at once, but not sure whats involved.)

but gsb also has additional fixed stop, and all the other exit types (like secondary filter reverse & many more)

What I think your mean by your statement "Does this mean that profit lock and atr stop, is now working in terms of logic in real trading?" is

does multiple active stops held in the market work in MC live trading. Answer is likely no. TS doesnt have this issue and work around are use market if touched order
or re-code the gsb systems to have only one active stop. While I appreciate you want this feature, there are issues that are higher priority. Sorry.
You could also chase MC over this as its a feature TS has that MC doesnt.

admin - 17-11-2023 at 12:02 AM

@daniel, the build 66.55 due to a bug uses plock exit for the fixed stop as well, so that will be nice for you on mc.
fixed in the next release though. You can look at the minor code changes between the two builds

admin - 20-11-2023 at 01:33 AM

next release, weekend only moc exit. Been a long time coming to get that


weekend.png - 72kB

admin - 28-11-2023 at 11:43 PM

Next build will have hull moving average and weighted moving average and a number of indicators that im testing and will release if I think they have merit.
Likely a week or so away

admin - 11-12-2023 at 05:02 AM

66.67 released. fix for expired code. Has new features still being tested

JozefSusko - 12-12-2023 at 08:26 AM

Hi Peter,

I'm missing the file: GSB_Scripts_2023_11_28
where can i download it?

thank you

Carl - 12-12-2023 at 11:19 AM

Quote: Originally posted by JozefSusko  
Hi Peter,
I'm missing the file: GSB_Scripts_2023_11_28
where can i download it?
thank you


Hi Josef,

I used the ResourceManager to update GSB.
The new scripts file is in the same GSB directory.

JozefSusko - 12-12-2023 at 11:37 AM

Quote: Originally posted by Carl  
Quote: Originally posted by JozefSusko  
Hi Peter,
I'm missing the file: GSB_Scripts_2023_11_28
where can i download it?
thank you


Hi Josef,

I used the ResourceManager to update GSB.
The new scripts file is in the same GSB directory.


Hi Carl,

yes i used, RM to update
but the last one is GSB_Scripts_2023_10_18

14_11_2023.JPG - 15kB

Carl - 12-12-2023 at 11:42 AM

Quote: Originally posted by JozefSusko  
Quote: Originally posted by Carl  
Quote: Originally posted by JozefSusko  
Hi Peter,
I'm missing the file: GSB_Scripts_2023_11_28
where can i download it?
thank you


Hi Josef,

I used the ResourceManager to update GSB.
The new scripts file is in the same GSB directory.


Hi Carl,

yes i used, RM to update
but the last one is GSB_Scripts_2023_10_18


Just checked, you are right.
Peter's new ELD does not include the function GSB_Scripts_2023_11_28.

Solution for now: create a new function with:
name = GSB_Scripts_2023_11_28
content = GSB_Scripts_2023_11_28 = 0

Ketil - 13-12-2023 at 02:05 AM

An easy solution is to put a comment on front of the line Value1 = GSB_Scripts_2023_11_28

Scripts.JPG - 9kB

admin - 13-12-2023 at 02:53 AM

Correct about the comment. This will be fixed next build. There is updated functions that are being tested by me, but not released so a comment is a good fix for now.

Build Alpha or GSB

MayNuts - 14-12-2023 at 08:18 AM

Hello, my name is Nikos and I am new to the forum. I m not sure if this the right place for my question so I apologize in advance if it's not.

I recently ran into the concept of developing trading strategies through software like Build Alpha or GSB. I'm about to choose which one I will buy and since most people in here seem to have good experience with both of them I would like to ask your opinion.

I m mostly swing trading and I try to improve my existing systems, plus check their robustness more accurately.

I would appreciate your insights, thank you in advance for your time.

admin - 14-12-2023 at 11:24 PM

Hi Nikos, welcome to the forum.
A number of users of GSB have BA, TSL & Adaptrade builder.

I have not looked at BA for over a year, so not sure how valid my comments are. However user feed back is positive, and I cant say that about all other competitors.
GSB has multiplication of processing and walk forward via as many cloud machines as you want to have. At times there is also considerable free cloud power. (depends on my useage)
BA did not have any WF tools at all, but it has other checks to see if a system is valid.
GSB has a very well explored methodology that you can look at 1000's of systems and see what sort of setup they collectively work best on. To my knowledge no other system builder has anything like this.
BA I think is long only or short only, and not both at the same time. While im increasingly a fan of long only systems, I could also argue long / short could be more robust as you have double the amount of trades. Feedback to me is GSB systems did better out of sample, but that could be to so many factors and market conditions & other variables
Examples of systems out of sample since June 2018 is https://trademaid.info/gsbhelp/GSB_AG_swingAG101ES.html (updated this week) Regardless of what product you choose, the biggest risk to me is systems failing and or poor market conditions

zykem - 15-12-2023 at 10:13 AM

Hi Nikos,

I have used BA before but was never able to produce any systems that performs out of sample. It seems to be a very basic program that creates simple systems.

There is a lot more to GSB that just putting technical indicators together. It is more complex and takes some time to learn but once you do that and learn the methodology to create system, you will see that it's far superior than any of these other system builders out there.

You will also find that this is an open forum where users and peter share their finding on markets, or if they find something new that works. There is nothing else out there that can compare. I have been a user since the beginning of 2020 and some of those early systems that I created still performs today without any reoptimizing or anything.

Wish you luck
Nick







Quote: Originally posted by MayNuts  
Hello, my name is Nikos and I am new to the forum. I m not sure if this the right place for my question so I apologize in advance if it's not.

I recently ran into the concept of developing trading strategies through software like Build Alpha or GSB. I'm about to choose which one I will buy and since most people in here seem to have good experience with both of them I would like to ask your opinion.

I m mostly swing trading and I try to improve my existing systems, plus check their robustness more accurately.

I would appreciate your insights, thank you in advance for your time.

admin - 15-12-2023 at 03:47 PM

Update on next GSB build
The next build has had some very difficulty bugs to fix in new features, and so has taken longer than expected.
One of the features is the ability to make daily bars out of intra day, which I specifically wanted as a filter on for the very successful long only NQ / ES systems.

A lot of time from me is also going into the new batch processor for TS. Thats getting close to a beta release.
https://trademaid.info/gsbhelp/BatchProcessorforTradestation...

Ketil - 16-12-2023 at 09:34 AM

Hi Nikos,

I have been using Build Alpha for some time, and I am very satisfied with it. Currently, I am trading 11 BA strategies, and all of them have been profitable this year. This is not the case with several of my other systems. Based on my experience, BA is a great tool for building simple and robust strategies. When I build systems with BA, I always start with a key idea that has a raw edge. I can provide an example: Day of week patterns offer an edge for many symbols. If you for example want to build long-only systems on KC, you go long on Monday and close at the next open. For short-only you go short on Friday and close Monday morning. If you want to build day of week systems for Si/GC/HG, you do the opposite, long on Friday and short om Monday. However, you can't buy and sell every Friday and Monday; you need a filter to ensure you only take trades under the most favourable conditions. To find these filters, I use BA. I have attached equity curves for some of these day of week systems. You can of course build other types of systems as well. If you want to build mean reversion systems for ES, for example, you can use conditions like RSI2<20 and let BA find filters. This approach will result in systems with good profitability on unseen data.

You're asking whether to choose GSB or BA. In my opinion, you need both. GSB is excellent and a must-have for all system builders. However, you also need BA because it allows you to build entirely different types of systems. In my view, the holy grail in trading is to trade a portfolio of uncorrelated systems. If you have both GSB and BA systems, it's easy to assemble such a portfolio.


KC_DOW_Long_short.PNG - 27kB

SI_DOW Long_short.JPG - 287kB HG_DOW Long_short.JPG - 356kB

admin - 17-12-2023 at 07:03 PM

66.69 released. This has new features (hidden short term) that are being tested.
New scripts also. You dont need new scripts apart for dummy filter until I unlock the new features

admin - 19-12-2023 at 03:56 AM

66.70 released. Has new features I am testing. New eld not need unless you want the dummy script
newest eld here, but its included in the update
GSB_ALL_FUNCTIONS_2023_12_18_1

Attachment: Login to view the details


admin - 27-12-2023 at 05:41 PM

Critical bug for TS users. Please look here, and post comments in that thread too.
https://trademaid.info/forum/viewthread.php?tid=270&page=11#...
Next build of GSB will have new settings for Swing23 hour NQ systems using the new Hull filter. Under testing now. SHould be released early next week
Has a fix for Ninja trader users too. (Not allow building of nt systems when ne doesnt have the indicators selected)

admin - 29-12-2023 at 01:02 AM

Just an update for the week.
There is a new build of GSB automation out, but not documented yet as more to be added. Has some of the new filters added.
One more filter to be added in the next build of GSB that I need to test before the next build is released.
Much of my time this week has been spent on the Batch processor for Tradestation to speed up the human time getting ts reports.
Happy new year to you all.
Peter

admin - 16-1-2024 at 11:35 PM

whats coming in a few builds
WF of filters.
option to wf stop and PT, (or not wf)
option to remove duplicate / redundant indicators.
There are some redundant indicators
IE loseLessPrevCloseD,-> CloseLessPrevCloseDBpv
& RoofingFilter1Pole -> RoofingFilter1Pole2
plus more in the pipeline.
With this setting to true, if you select true on one the old indicators, it will be set to false, but the other / primary version will be set to true.
the default for this new setting will be true
Filter daily simple moving average. THis did not work as well as hull moving average


2024-01-17_16-24-17.png - 407kB2024-01-17_16-32-06.png - 68kB2024-01-17_16-33-48.png - 85kB

admin - 17-1-2024 at 08:59 PM

66.82 released. New eld's are included.

Carl - 18-1-2024 at 04:03 AM

Hi Peter,

Just tried to get the updated 66.82 version by using ResourceManager, but got no downloaded files in the temp directory.

My version is still 66.69.

Is it me or a more general issue?

Thanks

admin - 18-1-2024 at 07:39 PM

Hi Carl
there was fa file copy on by back end, should be fixed now. thanks for letting me know. You were the second person to complain so i looked into it in more detail

Carl - 21-1-2024 at 02:34 PM

Hi Peter,

Hope you had a good weekend.

I have been using the latest GSB version 66.82 and noticed a few changes in the WF section.

1. Please tell me, what does "most common" mean in the "Run WF" menu?

2. After completion of the WF, I can see two separate WF blocks under tab "WalkForward". Underlying parameters? What does this mean?

Thanks


admin - 21-1-2024 at 03:52 PM

@carl,
most common will use the parameter that has been used the most common in previous runs. The feature was introduced in 2021
The logic was the most common parameter is often the best and most stable. I have not used this for a long time. Was fairly sure I documented this but cant find it in the docs.
can you show me a picture of 2 ?

Carl - 22-1-2024 at 01:48 AM

Quote: Originally posted by admin  
@carl,
most common will use the parameter that has been used the most common in previous runs. The feature was introduced in 2021
The logic was the most common parameter is often the best and most stable. I have not used this for a long time. Was fairly sure I documented this but cant find it in the docs.
can you show me a picture of 2 ?


Hi Peter,

"most common" (new in version 66.82 compared to version 66.69):
Common GSB 66.82 20240122.JPG - 62kB

"two WF blocks" (sometimes visible in version 66.82, but not always):
My tests were done in "alpha mode".
Two WF GSB 66.82 20240122.JPG - 267kB

and a third question: "additional value". What does this mean?
(again sometimes visible in version 66.82, but not always)
Parameter values GSB 66.82 20240122.JPG - 27kB

Thanks


admin - 22-1-2024 at 06:39 PM

thats related to internal logic by ga that the programmer used, but i did not understand. It should be removed in a later build

admin - 26-1-2024 at 05:51 PM

updates for the week.
We are looking into indicator correlation. There were some that looked like identical, but then we add more decimal places we see this is not the case and it gets more complex in that some things are really close but not the same. Also it was found that while indicators matched gsb, some had logical bugs. IE gsb_closePos
Note also how some things like many moving averages are nearly identical.
The end result is we will have the option for not having identical indicators, but we may also have a potential user adjustable threshold where you can choose the level at which high correlation indicators are not duplicated.

unnamed.png - 106kB

TradingPrice - 28-1-2024 at 12:12 AM

@Admin

Great work!

1- Could you please publish correlation for all indicators up to say 80%?

2- How about we change parameters for matching indicators (for example, Closetohighlow9 keep same parameter range, but for Closeothighlow5v4pos we chage the range.. more of like PercentR variants)

3- row 21: CloseLessHighDCloseD = CloseOverLowD ? that doesnt make sense

TradingPrice - 28-1-2024 at 12:15 AM

Quote: Originally posted by admin  
thats related to internal logic by ga that the programmer used, but i did not understand. It should be removed in a later build


So this additional value in brackets apparently affecting AI Walk Forward

Carl - 28-1-2024 at 01:49 AM

Hi Peter,

i agree some indicators generate the same outcome in GSB strategies.

Be careful using correlation. The overall correlation can be very high, but the correlation can change over time and have cycles with higher and lower values.

Another thought, (Close-average) is much more stationary compared to average alone.
Might be a good idea adding this kind of "averages" indicators to GSB to see if these have merit.

Thanks


admin - 28-1-2024 at 07:14 PM

IM happy to publish the entire indicator set correlation, however there are bugs that must first be fixed. ie close.pos and I suspect there are some that are identical kama2 vs adaptivemoving average, so these issues need to be fixed first.
Longer term im also entertaining the idea of
movingaverage(len,type)
if type=1 then result = average
if type=2 then result = xaverage
if type=3 then result = weightaverage etc

this would mean testing the top 10 indicator will give more true variation, and when we do a wf we should end up choosing the best type of average for the system


I like the idea of average-close (not close-average) and have implimented it a lot already, but open to expanding this down the track.

The addition of brackets do not affect wf results, they only tell more whats going on in the paramter array - which even I dont understand. Removed in next build regardless.


Not clear what you mean below
3- row 21: CloseLessHighDCloseD = CloseOverLowD ? that doesnt make sense


I dont want to have names that sound totally different for indicators that are the same


ie not
Closetohighlow9Lng
Closeothighlow5v4posSht


but rather
Closetohighlow9Lng
Closetohighlow9sht

TradingPrice - 1-2-2024 at 05:40 PM


Quote:

I dont want to have names that sound totally different for indicators that are the same


ie not
Closetohighlow9Lng
Closeothighlow5v4posSht


but rather
Closetohighlow9Lng
Closetohighlow9sht


Yeah, that would work :thumbup:




Quote:

Not clear what you mean below
3- row 21: CloseLessHighDCloseD = CloseOverLowD ? that doesn't make sense


I meant that reading the functions we have in Tradestation for both, it wouldn't seem that they'd be 99% corealted
one is Close - HighD + Close - CloseD
the other is Close over LowD


on another note, GSB_DecisionExit8 has case 1: IRevAndSFRev
which isnt available to select on GSB from the list on Exit Modes?




admin - 1-2-2024 at 05:47 PM

@tradingprice, I will put your comments on the to do list to look into.
We are still working through indicators looking for high corelation issues and bugs. Its very time consuming but will publish corleation table when we have resolved all issues. Likely a few weeks away is my guess. Too soon to tell now

admin - 1-2-2024 at 05:55 PM

@tradingprice, I see it in gsb


2024-02-02_10-53-21.png - 177kB

TradingPrice - 1-2-2024 at 06:06 PM

Quote: Originally posted by admin  
@tradingprice, I see it in gsb


That's case #2 : IFlatRevAndSFFlatRev


admin - 9-2-2024 at 01:15 AM

This is the correlation with remove duplicate indicators on true.
all 80 indicators for GSB purchased users, which I think dropped to about 76 after duplicates was set to true
Interesting observation, things like moving averages are highly correlated before normalization. (adjusted into +100 to - 100 range) but after normalization are quite different

More work is going to be done on this over time.
I want to export the entire grid in csv format for users.

Attachment: Login to view the details

 Pages:  1  ..  10    12    14