136 Bug - ALL SeqOpt results are different than the Original + SeqOpt results after being retested.

136 Bug - ALL SeqOpt results are different than the Original + SeqOpt results after being retested.


The Original strategy is taken into SeqOpt, then SeqOpt will change the parameters of the strategies by the stable region found out from the settings your chose,
You are now able to see both of the results inside that single strategy file,
• Results of the Original strategy (before the SeqOpt)
• Results of the Optimized strategy (after the SeqOpt)

Its all great until you retest that same strategy,
when you retest the same strategy using the same data and periods etc, you will see different results!, not sure why (verified that and made sure a couple of times, even restarted SQX and cleared the cache).


Attachments
No attachments
  • Votes +2
  • Project StrategyQuant X
  • Type Bug
  • Status Fixed
  • Priority Normal

History

k
#1

Karish

24.12.2022 14:27

Task created

k
#2

Karish

24.12.2022 14:29
Voted for this task.
k
#3

Karish

24.12.2022 18:10

I'm not sure what's going on with it,

like 90% of all strategies that passes SeqOpt and auto-applied the new parameters shows one result when choosing the SeqOpt view,

but when those results are getting retested on the same data period spread etc, those same results with those newly applied parameters shows different result..,


Diffidently a SeqOpt bug.

E
#4

Emmanuel

24.12.2022 19:18
Voted for this task.
k
#5

Karish

11.02.2023 23:19

My theory from experience with WF and SPP tests leads me to the following reason why that thing happens,

It might be that SeqOpt / SPP / WF/M use Genetic Optimization rather than Exhaustive Optimization method,

even when we clearly specify the % limitations of +- and the number of steps etc, which is obviously meaning that we do want to limit our optimizations,

still under the hood SQX uses the Genetic Optimization method so every-time that the user wants to use SeqOpt / SPP / WF/M tests the user will get different results from his previous results with the same exact dates, data, settings etc..


The genetic optimization MUST BE OPTIONAL!


this was already mentioned here:
https://roadmap.strategyquant.com/tasks/sq4_9802


and here by @Bentra
https://roadmap.strategyquant.com/tasks/sq4_8538


This must be fixed..!

g
#6

Lee Guan Chuan

13.02.2023 07:00

Status changed from New to Fixed

k
#7

Karish

13.02.2023 10:54
Thanks!, Cant wait to test it ;)
k
#8

Karish

13.02.2023 13:15
@Lee Guan Chuan

Hii it is also very important to fix this issue with SPP, WF, WFM

because the genetic method is also activated there rather than a standard Exhaustive Optimization method,

at least give the user the option to choose between the 2 methods.


thanks.

MF
#9

Mark Fric

01.03.2023 13:23
behavior explained in the last comment here: https://roadmap.strategyquant.com/tasks/sq4_9922




Votes: +2

Drop files to upload

or

choose files

Max size: 5MB

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

...
Wait please