The problem with the 50s
Added 2022-11-20 18:06:16 +0000 UTCTwo weeks ago, a reddit user pointed out an issue with Tillerinobot that started when the pp formula was changed to punish 50s directly (not just indirectly through accuracy). I had known about this for a while, but hadn't done anything about it yet. I had done some research (see response), but got distracted with other things. So these past days, I got off my ass and finally finalized my approach.
The fix went live today and I wrote all about my approach over on reddit. If you're into this kind of stuff, you can find the Jupyter Notebook attached below that contains all the code for the fitting of the polynomials with boundary conditions.
