Detect similiar strategies

It would be excellent to have a mechanism which detects similiar strategies during buliding process  . 
Here is a mail from adaptrade i recieved today 


As another way to maintain and potentially increase population diversity, I've added an anti-cloning rule, which identifies duplicate strategies at the end of each generation and replaces them with new, randomly generated ones. Even after the results start to converge, this prevents duplicate strategies from dominating the population.


Also, five new mutation operators have been added: subtree substitution, subtree simplification, subtree complication, subtree addition, and subtree pruning. The choice of mutation operator depends on probabilities that are self-adapted by the build process. In other words, the probability of selecting any given mutation operator is evolved during the build process based on how effective each operator is. These probabilities can change over time depending on the evolutionary process and how effective each operator is at any given point during the build.



Clnx

Attachments
No attachments
  • Votes 0
  • Project StrategyQuant X
  • Type Feature
  • Status Fixed
  • Priority Normal

History

?
#1

anonymous

28.08.2017 10:40

Task created

MF
#2

Mark Fric

28.08.2017 12:11

Status changed from New to Confirmed

this is already working in SQ3 and it will be added also into SQ4. 
?
#3

anonymous

28.08.2017 13:20
Hi. it is usual problem that sq3 gives me lot of very similiar strategies. If this is alredy implemented still it doesnt makes what we need. Then maybe would be good to have and option to dtct ths statgs fm databank
?
#4

anonymous

29.08.2017 00:14
I just wanted to add this as a feature request as well! I also cannot confirm that this was implemented (or at least it then was not working) in SQ3, I always had and have TONS of duplicate strategies within the population,

with just slightly varied parameters. This should be a lot more effective for SQ4, just like Adaptrade Builder does it (and it works great there by the way).

IH
#5

clonex / Ivan Hudec

23.05.2018 12:10
Terrible. In RC2 I have 500 "same" strategies in the databank. This is something which should be solved with priority
c
#6

cullo6

23.05.2018 13:00
I know that here is a lot of work on other things, but I think that this should have higher priority. 5 000 strategies in databank is good, but not if they are the same or similiar. Hours of building are after that useless, when other good strategies are thrown away due to a full databank.
https://roadmap.strategyquant.com/tasks/sq4_1285
https://roadmap.strategyquant.com/tasks/sq4_1519
h
#7

hankeys

24.05.2018 08:16
určitě vyřešit co nejdříve, genetika se tímto stává téměř nepoužitelnou


je to o dost horší než v SQ3, kde taky genetiky generuje spustu duplicit, ale tam asi maže úplné duplicity, protože sem myslím neviděl strategii, která by měla stejný net profit, což u SQ4 neplatí...tady je to duplicitní naprosto

MF
#8

Mark Fric

25.05.2018 14:23

Status changed from Confirmed to Fixed

added filtering for similar strategies, I created a separate task to improve diversity during genetic evolution: https://roadmap.strategyquant.com/tasks/sq4_1684

Votes: 0

Drop files to upload

or

choose files

Max size: 5MB

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

...
Wait please