118 - Builder Island 1 - different fitness in graph and in list of last generation for 1st island (if is issue real - check stagnation calculation)

118 - Builder Island 1 fitness graph + Last genaration list
See attached pictures and config.
I have unchecked "Detect same strategies in population and replace them with newly generated ones" in genetic options in "Fresh blood" - Last generation for island 1 is more stable for comparison of fitness..
And I have checked "Show last generation databank for island 1 only".
The fitness in fitness graph for island 1 (I do not know if is in graph IS fitness???) is different then in Last generation list for island 1 in databank for all fitness and IS fitness only too. See average for best 3 strategies mainly (0,76 in fitness graph x whole fitness IS+OOS 0,6566666 and IS fitness only 0,69666).

Where is problem? Why is fitness graph different form Last generation list fitness?

Attachments
2PC00120 EURUSDM30 LIMIT.cfx
(21.75 KiB)
SQX118 Builder fitness graph vs IS fitness in last generation list different fitness 2 fitness IS.gif
(319.30 KiB)
SQX118 Builder fitness graph vs fitness ISanOOS in last generation list different fitness.gif
(301.49 KiB)
  • Votes +2
  • Project StrategyQuant X
  • Type Bug
  • Status Fixed
  • Priority Normal

History

t
#1

todtv

25.02.2019 19:52

Task created

t
#2

todtv

25.02.2019 20:05
Voted for this task.
t
#3

todtv

25.02.2019 20:12

Attachment 2PC00120b EURUSDM30 LIMIT.cfx added

Attachment SQX118 Builder fitness graph vs fitness last generation list different fitness on 1st position.gif added

I see the problem is often in fitness difference at 1st place too.

See attached file - difference graph 0,84 x last gen. list 0,66.


Either I do not understand what is in fitness graph and fitnes in Last genration list for 1st island or it is something somewhere bad.

t
#4

todtv

25.02.2019 20:20

Attachment log_2019_02_25.zip added

I add ziped log file too.
h
#5

hankeys

25.02.2019 20:21
Voted for this task.
t
#6

todtv

25.02.2019 20:30

It will be good to check which fitness calculations are used for stagnation calculations too (if it is somewhere issue in fitness in graph or last generation fitness).


For example due this betatest check is usable for my to see fitness information for some island and list of last island population too.

t
#7

todtv

25.02.2019 20:33

Subject changed from 118 - Builder Island 1 - different fitness in graph and in list of last generation for 1st island to 118 - Builder Island 1 - different fitness in graph and in list of last generation for 1st island (if is issue real - check stagnation calculation)

EDIT: I use weighted fitness in ranking option (here for example R/DDratio w2 + SQN w 1). May be in this is problem - if we use some kind of weighted fitness the other fitness calculation for graph stagnation etc.?? are by fitness return / DD ration only not by weighted fitness.


Stagnation calculation must use fitness which is choosed in Ranking option. It is good to check if it use choosed weighted fitness or it use wrongly some kind of simple fitness only (R/DD ration only, Net profit only etc.) even if some weighted fitness is set in Ranking option.

t
#8

todtv

25.02.2019 21:34

Last thinking about it from my side after observing last generation list:

Really does genetic evolution use the set weighted fitness by Ranking option or generally set fitness in Ranking option setting?


I do not have more intermediate results for observation of detalis in calculations and I can only estimate where is problem.

TT
#9

Tamas

21.04.2020 09:01

Status changed from New to Fixed


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