NNUE reverse engineered, an ongoing study

[chrisw]

NNUE reverse engineered, an ongoing study

What is a chess piece really worth?
Chess System Tal 3 judges positions with a neural network that holds no piece-value table at all — so we reverse-engineered one. By regressing millions of the engine's own evaluations we recovered what each piece is truly worth to it, how those values slide from a crowded opening to a bare endgame, and the piece-square maps it taught itself with nothing hand-coded: knights that crave outposts, a king that flips from cowering in the corner to marching up the board, and a queen worth nearly nine pawns when there's a full board to attack.

Read the study here: https://whittingtonchess.com/reverse-engineered-nnue

WIP, comments, critiques welcomed. Authors who might like their NNUE's reverse engineered and analysed, get in touch.

[royb]

I found this fascinating! We all know that the value of any piece changes as the position on the board changes. What I had not considered nearly as much was the piece value changes as the phase of the game changes also. Looking forward to your analysis of Stockfish 18!

[chrisw]

Stockfish 18 reverse engineered analysis posted

https://whittingtonchess.com/reverse-engineered-nnue