NN-longtime testrun of Stockfish 201225 vs Lc0 0.26.3 66680 finished ("SuFi for the poor").
https://www.sp-cc.de
(Perhaps you have to clear your browsercache or reload the website)
SPCC: Longtime-testrun of Stockfish vs Lc0 66680 finished
Moderators: hgm, Rebel, chrisw
-
- Posts: 2435
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
-
- Posts: 253
- Joined: Mon Nov 16, 2020 12:13 pm
- Full name: Manuel Rivera
Re: SPCC: Longtime-testrun of Stockfish vs Lc0 66680 finished
Really nice. Thanks for your work !
Raspberry Pi4 bot : https://lichess.org/@/BetterAnalyze
-
- Posts: 2435
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: SPCC: Longtime-testrun of Stockfish vs Lc0 66680 finished
My pleasure.
The result is the best of all longtime testruns so far: Lc0 -50 Elo behind latest SF-Dev. In a TCEC Superfinal with really long thinkingtime and extremly fast hardware, the result could be (and I expect that) much closer. But I have no doubt, that Stockfish is still stronger. When a Leela-Ratio near 1.0 is used, only, of course.
But Lc0 has made much progress. Without the nnue-revolution, Stockfish would be clearly weaker and Lc0 would be the world's #1 engine.
-
- Posts: 253
- Joined: Mon Nov 16, 2020 12:13 pm
- Full name: Manuel Rivera
Re: SPCC: Longtime-testrun of Stockfish vs Lc0 66680 finished
Yeah i agree with all you said.
Raspberry Pi4 bot : https://lichess.org/@/BetterAnalyze
-
- Posts: 3019
- Joined: Wed Mar 08, 2006 9:57 pm
- Location: Rio de Janeiro, Brazil
Re: SPCC: Longtime-testrun of Stockfish vs Lc0 66680 finished
I will be curious to see how the same net fares inside Ceres.pohl4711 wrote: ↑Thu Dec 31, 2020 3:37 pmMy pleasure.
The result is the best of all longtime testruns so far: Lc0 -50 Elo behind latest SF-Dev. In a TCEC Superfinal with really long thinkingtime and extremly fast hardware, the result could be (and I expect that) much closer. But I have no doubt, that Stockfish is still stronger. When a Leela-Ratio near 1.0 is used, only, of course.
But Lc0 has made much progress. Without the nnue-revolution, Stockfish would be clearly weaker and Lc0 would be the world's #1 engine.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
-
- Posts: 2435
- Joined: Sat Sep 03, 2011 7:25 am
- Location: Berlin, Germany
- Full name: Stefan Pohl
Re: SPCC: Longtime-testrun of Stockfish vs Lc0 66680 finished
Me, too. When somebody builds a working bmi2 (intel CPU)/CUDA 11 (RTX 2060)-binary of Ceres, I will do a testrun very soon.Albert Silver wrote: ↑Sat Jan 02, 2021 4:01 amI will be curious to see how the same net fares inside Ceres.pohl4711 wrote: ↑Thu Dec 31, 2020 3:37 pmMy pleasure.
The result is the best of all longtime testruns so far: Lc0 -50 Elo behind latest SF-Dev. In a TCEC Superfinal with really long thinkingtime and extremly fast hardware, the result could be (and I expect that) much closer. But I have no doubt, that Stockfish is still stronger. When a Leela-Ratio near 1.0 is used, only, of course.
But Lc0 has made much progress. Without the nnue-revolution, Stockfish would be clearly weaker and Lc0 would be the world's #1 engine.