Dual NNUE-new leaf in Stockfish dev

[Werewolf]

To my surprise this dualnet SF is worse than SF16 in all testsuites I tried! And also worse mate solver . 31.12. version was very good solver.

Are you sure the search hasn't been changed?

[Jouni]

Github says: Showing 12 changed files with 297 additions and 198 deletions. Search.cpp is not there.

[chessica]

could it be that test suite performance is completely irrelevant to actual engine dev? shocked

Yes, it is absolutely the wrong way to solve a chess position on the board, it is better to just roll the dice!

(I don't even know if that material imbalance table is still in Stockfish, the PSQT tables I think have gone, sorry for not being up to date with it. I can't find anything about material.cpp anymore in the source so it is I think long gone...)

all classical eval in stockfish has been removed, it's nnue all the way

[Eduard]

I'm not happy with the current parameters for two networks. Example: the following two positions are solved immediately with the previous engine with 1 network. With two networks, SF dev takes 30 to 60s on 6 cores.

Pos 1

r1n2N1k/2n2K1p/3pp3/5Pp1/b5R1/8/1PPP4/8 w - - 0 1


Pos 2

2r2r1k/qb3p1p/p2p1Np1/np1Pp1P1/2P5/3RP2P/5R2/4Q1K1 w - - 0 1

For me personally, I changed evaluate.cpp so that it be solved immediately. I don't care whether I have -2 Elo in Bullet. For me, analysis comes first, and this is where Stockfish has become weaker now.

[Ciekce]

could it be that test suite performance is completely irrelevant to actual engine dev? shocked

Yes, it is absolutely the wrong way to solve a chess position on the board, it is better to just roll the dice!

stockfish exists to be the strongest chess engine in the world in actual games

its usefulness as an analysis tool is secondary to that