You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I just ran a genetic evolution on my strategy for 24 hours straight. It was with 200
EPOCHS = 200
POP_SIZE = 5
NBLOCALE = 3
But the results I get are confusing. On the webserver results page at localhost:8182, it spit out 5 toml settings for me, as follows:
My concern is that 3 out of the 5 results are identical. What is the reasoning behind showing the same thing 3 times? In addition, how did it pick these 5, because the logs folder shows this:
There are 23 toml files here. Which ones am I supposed to use, and how come the results page does not show 23, but rather 5 results? And, I set it to use 3 locales, but it used 9, why?
Confusingly, the parameters selected on the results page don't seem to be the best settings, because if I read the japonicus.log file, I can scroll through and see this:
Relative profit on evolution dataset: 170457.160
Current parameters are viable.
Relative profit on evaluation dataset:
Test on random candles...
Relative Profit: 15698.371
# of Trades: 21.0
Sharpe: 6.76
170457.160 final profit ~~~~
[fastMA]
optInTimePeriod = 69.32
[gekkoRSILength]
interval = 20.6
redacted
This clearly shows that japonicus found better settings, but did not report them on the results page. So, which results should I trust, and is this a bug that the results page only displayed 5 out of 23 toml settings?
Lastly, if some settings are twice or 10x as profitable as another, or duplicates of another, why does it even bother outputting these toml files?
I know it was a lot of questions, but this will help me understand the system better.
Thanks!!
The text was updated successfully, but these errors were encountered:
IAMtheIAM
changed the title
Question about intepreting results
Question about intepreting results and possible webserver bug
Jul 11, 2019
I can confirm that this is confusing.
And then I add one more question, what does "Evaluation Break" show on the UI? There is one blue line, "Evaluation", and one "Secondary". I don't know how to interpret this.
Hey guys, (sorry for the delay @IAMtheIAM ). Japonicus is a somewhat a complex response for the insanely complicated problem of candlestick trading strategy fitness XD (it is not a definitive answer tho).
So I've just updated the wiki (Results section) to address your questions... hope it is a bit clearer now.
I just ran a genetic evolution on my strategy for 24 hours straight. It was with 200
EPOCHS = 200
POP_SIZE = 5
NBLOCALE = 3
But the results I get are confusing. On the webserver results page at localhost:8182, it spit out 5 toml settings for me, as follows:
My concern is that 3 out of the 5 results are identical. What is the reasoning behind showing the same thing 3 times? In addition, how did it pick these 5, because the logs folder shows this:
There are 23 toml files here. Which ones am I supposed to use, and how come the results page does not show 23, but rather 5 results? And, I set it to use 3 locales, but it used 9, why?
Confusingly, the parameters selected on the results page don't seem to be the best settings, because if I read the japonicus.log file, I can scroll through and see this:
and also this
This clearly shows that japonicus found better settings, but did not report them on the results page. So, which results should I trust, and is this a bug that the results page only displayed 5 out of 23 toml settings?
Lastly, if some settings are twice or 10x as profitable as another, or duplicates of another, why does it even bother outputting these toml files?
I know it was a lot of questions, but this will help me understand the system better.
Thanks!!
The text was updated successfully, but these errors were encountered: