When looking for minimized fitness, score sum of all chromosomes should maintain the same when inverting the scores, so that the normalized fitness can work.
Particularly, if the sum becomes smaller after inverting scores, then on line 206 there is a chance that level might never reach point, if point happens to be greater than the new sum.
I am using another method of inverting the score which maintains the sum of scores at the same time.
preserve sum when inverting scores
record best solution and number of stable generations
added chance to choose second parent based on fitness
Oops I pushed a couple of extra commits to my master which don't belong to this pull request... Opened another request. Sorry :)