[Build 136] not respecting custom parameter levels

The level is set for 0 for every single set yet strategies still come with non 0 Levels for some indicators.
Attachments
Screenshot_423.png
(19.38 KiB)
Screenshot_425.png
(38.90 KiB)
  • Votes +4
  • Project StrategyQuant X
  • Type Bug
  • Status In progress
  • Priority Normal

History

b
#1

bentra

19.12.2022 02:33

Task created

b
#2

bentra

19.12.2022 06:00

Subject changed from 136 RC 4 not respecting weight of 0 for the random set to 136 RC 4 not respecting Disparity Index custom parameter level

Description changed:

The level is set for 0 for every single set yet strategies still come with non 0 Levels for disparity index.

Attachment Strategy 31768.sqx added

Attachment bugbuild.cfx added

Strategy 31768.sqx
(34.15 KiB)
bugbuild.cfx
(46.68 KiB)
b
#3

bentra

19.12.2022 06:01
Voted for this task.
MF
#4

Mark Fric

19.12.2022 09:27

Status changed from New to In progress

can you attach Java snippet at least for your DisparityIndexCrossesAboveLevel indy so I can reproduce it?
b
#5

bentra

19.12.2022 14:45

Attachment disp-SQExtension.sxp added

disp-SQExtension.sxp
(5.88 KiB)
E
#6

Emmanuel

19.12.2022 20:23
Voted for this task.
b
#7

bentra

21.12.2022 02:51

Subject changed from [Build 136] 136 RC 4 not respecting Disparity Index custom parameter level to [Build 136 Final] not respecting custom parameter levels

Description changed:

The level is set for 0 for every single set yet strategies still come with non 0 Levels for some indicators.

b
#8

bentra

21.12.2022 02:52

Attachment Screenshot_436.png added

Attachment Screenshot_435.png added

Attachment Screenshot_434.png added

Screenshot_436.png
(161.93 KiB)
Screenshot_435.png
(224.17 KiB)
Screenshot_434.png
(177.80 KiB)
Also happens for ROC. Same build config....
MF
#9

Mark Fric

22.12.2022 09:29
I still cannot reproduce it - you use custom template and fitness.


But I assume it happens because of genetic operations - we do have controls after the genetics, but they don't check the build settings of the blocks 100%.

So something like this could happen. 


Does it happen to you also if you switch to Random generation? It shouldn't.


We can add 100% checks in the next build, but it could potentially slow down building process somewhat.


b
#10

bentra

22.12.2022 16:47
It's rare. 

Is it happening during the crossovers? As I understand from what you told me before any single variable can come from any position from one parent and be inserted in to any position of the other parent. For instance the SL multiplier could become the ROC greater than level. But does that really make sense? What if only entire indicators could be swapped between parents. It would solve this problem in this ticket and probably make crossovers more beneficial...


k
#11

Karish

24.12.2022 14:30
Voted for this task.
b
#12

bentra

26.12.2022 00:55
If I'm right then this will fix this problem:


https://roadmap.strategyquant.com/tasks/sq4_9808

CG
#13

Chris G

09.04.2023 17:27
Voted for this task.

Votes: +4

Drop files to upload

or

choose files

Max size: 5MB

Not allowed: exe, msi, application, reg, php, js, htaccess, htpasswd, gitignore

...
Wait please