While being a decent game, it would probably be more fun if the focus shifted more towards adjusting parameters to make the machine learning work. For example, reduce the cost of later upgrades, add an adjustable 'learning rate' parameter, ability to focus on training specific skills, custom tracks to help training specific skills....
While writing this, my bikes learned themselves to keep making donuts in the first corner :D
Somehow my bikes no longer know how to steer after switching from lvl 5 to 6. They can't even complete any of the earlier levels anymore... Even worse, it caused me to get stuck in the 'steer wrong way' loop.
Got bored after my third puzzle. The amount of simple counting is too much compared to amount of thinking it requires, especially on the larger boards. Also, the game doesn't check if a puzzle has multiple solutions, which is highly unsatisfying.
The safe buy slider should be exponential, not linear (there is barely any difference between 98% and 99% and it's hard to select the 1/100th of the slider you need). I want to set it to 1e4 times lower than my current cash, or probably even 1e5 or 1e6 times lower. An auto-buy button to safe-buy all upgrades once would be nice too.
Thank you very much for your suggestions. I added the exponential slider for the safe-buy button to my todo-list. Regarding the auto-buy button: Simply a button which automatically performs one purchase operation on every upgrade depending and taking your current setting (x1/x5/x25/x100/max./safe) into account?
Dear Konnectore, thank you for your support. But now I wish I had 1000 "First Names". Sincerely Corporal Tutnix.