Subject: Stockfish 15.1 Released. Mon Dec 05, 2022 11:34 am
Elo gain and competition results With this release, version 5 of the NNUE neural net architecture has been introduced, and the training data has been extended to include Fischer random chess (FRC) positions. As a result, Elo gains are largest for FRC, reaching up to 50 Elo for doubly randomized FRC (DFRC). More importantly, also for standard chess this release progressed and will win two times more game pairs than it loses against Stockfish 15. Stockfish continues to win in a dominating way all chess engine tournaments, including the TCEC Superfinal, Cup, FRC, DFRC, and Swiss as well as the CCC Bullet, Blitz, and Rapid events.
Ghppn and Holofernes like this post
Chris Whittington
Posts : 1254 Join date : 2020-11-17 Location : France
I wish it was +25 Elo. But it did make very small progress at longer time controls and more cores. And that is a better change then we have seen in the last 2 years.
I wish it was +25 Elo. But it did make very small progress at longer time controls and more cores. And that is a better change then we have seen in the last 2 years.
I wish it was +25 Elo. But it did make very small progress at longer time controls and more cores. And that is a better change then we have seen in the last 2 years.
When did Stockfish last make a big/biggish gain?
From Stockfish 11 to Stockfish 12. Then it has been hard from them to gain much after NNUE was implemented. Outside of bullet and micro bullet time control testing.
I wish it was +25 Elo. But it did make very small progress at longer time controls and more cores. And that is a better change then we have seen in the last 2 years.
When did Stockfish last make a big/biggish gain?
From Stockfish 11 to Stockfish 12. Then it has been hard from them to gain much after NNUE was implemented. Outside of bullet and micro bullet time control testing.
I am looking at the discussion topic from the chief developer Vondele from July last year. Yes, it is not yet 2½ years. But that was of course a quantum leap. People get spoiled Vondele explains it in the graph as mostly speed gain. You see less Elo the higher the time control (I assume), because speed does not matter much there. Holes do. That there was still much gain after that is well more speed up too but with newer architecture speed went down again in the run up to Stockfish 14 (I'm just reading that graph now, I don't know much else about it).
For chess, the NNUE also meant, Uly wrote about this a bit once, more openings became playable. That is a huge benefit too. (Less holes in chess knowledge).
Home 
Vondele explains it in the graph as mostly speed gain. You see less Elo the higher the time control (I assume), because speed does not matter much there. Holes do. That there was still much gain after that is well more speed up too but with newer architecture speed went down again in the run up to Stockfish 14 (I'm just reading that graph now, I don't know much else about it).