And then you play a move that results in minus 50 points…
After every move, katago evaluates how many points it gains compared to passing, and reports this to both players.
sounds like useful idea for post game AI review
current “points compared to bot move” is not very objective metric
If your move is worse than pass, then its mistake indeed
Or performance index: (AI score after your move - AI score after passing)/(AI score after AI’s move - AI score after passing) x 100.
Performance 100 means your move is as good as AI. Performance 0 means your move is as bad as passing.
Do you just cap it at 100? I imagine near the end of a game, if I self atari and the bot would’ve just played some dame, maybe ai score after ai move and after passing could be quite small.
Could be like 18/0.1 *100 =18,000
It could be more than 100 if you find a better move than AI (this rarely happens but is possible), or negative if your move is worse than passing (e.g. self-atari). If passing or a dame is the bot’s top move than the formula can’t be computed (division by zero).