Standard way to compare Stockfish evaluation scores to Leela evaluation scores? | Stockfish | Page 1

exotic raft Jun 13, 2023, 1:17 PM

#

Hi everyone 🙂

I'm wondering if there is a "standard" or established way to compare the position evaluation of Stockfish to the position evaluation of Leela? As far as I understand, the two CP scores from Stockfish and Leela do have different ranges and interpretations (for example +1 pawn for Stockfish does not equal +1 pawn for Leela) so it's not that straight-forward. Is it better to compare the two WDL values?

Thanks for your help!

lapis iris Jun 13, 2023, 1:17 PM

#

they are both normalized by wdl model nowadays

#

so in fact +1 of stockfish should mean the same thing as +1 of leela

#

on average

exotic raft Jun 13, 2023, 1:26 PM

#

Thank you ver much for your answer! 🙏 I've seen that this is the case for Stockfish here: https://github.com/official-stockfish/Stockfish/wiki/Stockfish-FAQ#interpretation-of-the-stockfish-evaluation

Do you have by any chance a source that Leela does this as well? I wasn't able to find anything.

devout bane Jun 13, 2023, 1:26 PM

#

For Leela you need to set ScoreType to WDL_mu, then it'll use the same thing that Stockfish uses

#

You can ask more about it in Leela Discord

#Standard way to compare Stockfish evaluation scores to Leela evaluation scores?