Oh and by the way....increasing max speed shouldn't affect minimum speed and vice versa. Why don't you try minimum speed (changes just the min), max speed (changes just the max, acceleration/brake (changes how quickly AI can change its speed), and handling (changes how quickly the bike can turn relative to its speed).
This "game" is an example of bad machine learning....
You don't just randomly change things. Yes one of the first processes in learning is to figure out something that improves your performance, but after that, you must remember the opposite thing - what not to do, as well as determine how much of the change is beneficial, before it starts doing more harm.
Even 150 generations in, some of the AIs still can't figure out how to get past the straightaway immediately following the start line. This whole machine learning would become so much more efficient if the computer recognized things that it doesn't need to repeat.
Well I reviewed quite a few issues I have with the plot...but it's too long for the comment so the comment is in the link below:
https://textsaver.flap.tv/lists/22f9
Hi, interesting questions :) You can read the answers to all of them here..........................: https://textsaver.flap.tv/lists/22gg ..................................
This should be a good read for anyone who's interested more deeply in the lore of Nebezem.
(Note: I corrected the main hero's name in your questions where you accidentally referred to him as Magnus, instead of Marcus.)
 "Idle Machine Learning"){kind=icon}
 "ZS Dead Detective - Murder Case"){kind=icon}
 "Tough Growth"){kind=icon}
 "Tales of Nebezem: Four Fates"){kind=icon}
 "Sonny"){kind=icon}
 "The Last Stand: Dead Zone"){kind=icon}
 "Medieval Cop -The Invidia Game - Part 2"){kind=icon}
Good advice!