Closed olypnsf closed 1 year ago
Run lichess-bot and when this happens again, close it and then upload the files inside the lichess_bot_auto_logs
directory. There should be 2 files named recent.log
and old.log
.
I think he wants the engine to play stronger/better moves when it reaches +-1 eval. But I don't think lichess bot determines this. Unless you want to purposefully weaken the bot until it reaches +-1 Eval?
An evaluation of +1, which means the engine is about one pawn ahead, is not that great of an advantage. The evaluation is also only an estimate.
Chess engines are limited in how far ahead they can search. During the next move, the engine will be able to search two moves further than the previously (after the engine's move and the opponent's move), which may change the next evaluation due to getting more information.
It's possible that the +1 eval is from chessdb (possibly at depth 30+) and when the engine starts to think it doesn't see that far so it reports a lower score. Looking at a game and at your repo try disabling go_commands
, lowering Move Overhead
in uci_options
, and slightly increasing max_depth
in the opening book section. Also, I'm not sure how strong Cerebellum is at the moment. I think that the last version was released about 2 years ago.
Ok @AttackingOrDefending If i increase the max depth what is the best values?
Since you mostly play blitz, I would try increasing to 50 so that the bot gets as many moves from the opening book as possible. If that doesn't work then maybe try decreasing it to 12. You can set Move Overhead
to 500.
@olypnsf Are there any more questions for this issue?
When at this eval maybe like +1, It suddenly drops again. Which lead me confused why this engine don't convert their advantage clearly
I like to add some function for the engine in config.py if the eval is +1 then the engine is maximized to find the best move in the position