If using weighted fitness, and an already high fitness initial population ("Use strategies from initial population"), I´ve discovered that the genetic evolution
will not continue and get stuck forever trying to generate better
strategies if all the strategies in the results(!) database are "better"
than in the current generation of the genetic evolution.
I thought that the genetic evolution only
looks at its current own database of strategies for the current
generation and continues independently from the Results database, and if
it finds no better ones in the current generation than in the results
database, it will simply not copy them over to the results database and
continue with the next generation.
But that is not the case, it will continue generating strategies for the
current generation forever, as you can see in the screenshot (2800 of
2000 that are supposed to be in that generation and each new one gets
rejected with "dismissed, all strategies in databank are better"). Only
when I delete strategies from the results database (the worst ones), the
genetic evolution will go on to the next generation. If I don´t do
anything, it gets stuck forever in the current generation. The only
other solution is to make the results database much bigger, which wastes
RAM for nothing.
In my opinion, the genetic evolution should never be affected by the
results database. It should go on independently and if no better
strategies are found in the current generation than there are in the
results database, it should simply go on with the next generation
instead of trying to generate better ones than in the results database
for the current generation. Deleting strategies from the Results
database shows, that the genetic evolution is very well able to find
better ones in the next generation again, instead of trying to "force"
this on the current generation.