924 Partien gespielt / Turnier beendet Name des Turniers: Ryzen7 5700 Ort/ Land: LAPTOP-JTQQP8M7, Deutschland Spielstufe: Turnier 40/5 Hardware: AMD Ryzen 7 5700U with Radeon Graphics mit 4.096 MB Speicher
weird result, not only does dragon usually beat weaker engines, it also majorly underperformed here. Same with SF15.1 being below ethereal.
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:02 pm
Here is my analysis
Enter New Game Line 0.0 4r3/pp1b1pkp/3p1np1/4p3/6P1/qNP1QP2/P1P4R/1K5R w - - 0 1
Analysis by Lc0 v0.30.0-rc1:
24.Qh6+ Kg8 White is slightly better. +/= (0.63) Depth: 1/2 00:00:02 2Nodes 24.Qh6+ Kg8 25.c4 White is slightly better. +/= (0.59) Depth: 2/3 00:00:02 3Nodes 24.Qh6+ Kg8 25.c4 Qb4 White is slightly better. +/= (0.38) Depth: 2/4 00:00:02 4Nodes 24.Qh6+ Kg8 25.g5 Nd5 26.Qxh7+ White has an edge. = (0.30) Depth: 3/5 00:00:02 7Nodes 24.Qh6+ Kg8 25.g5 Nd5 26.Qxh7+ Kf8 The position is equal. = (0.20) Depth: 3/6 00:00:02 14Nodes 24.Qh6+ Kg8 25.c4 Re6 26.g5 Nh5 27.Rxh5 gxh5 White is slightly better. +/= (0.40) Depth: 4/8 00:00:02 35Nodes 24.Qh6+ Kg8 25.c4 Re6 26.g5 Nh5 27.Rxh5 gxh5 28.Qxh5 White is slightly better. +/= (0.44) Depth: 4/9 00:00:02 43Nodes 24.Qh6+ Kg8 25.c4 Re6 26.g5 Nh5 27.Rxh5 gxh5 28.Qxh5 Kf8 White is slightly better. +/= (0.43) Depth: 4/10 00:00:02 44Nodes 24.Qh6+ Kg8 25.c4 Re6 26.g5 Nh5 27.Rxh5 gxh5 28.Qxh5 Kf8 29.Qxh7 White is slightly better. +/= (0.40) Depth: 4/11 00:00:02 66Nodes 24.Qh6+ Kg8 25.c4 Re6 26.g5 Nh5 27.Rxh5 gxh5 28.Qxh5 Kf8 29.Qxh7 Qb4 White is slightly better. +/= (0.43) Depth: 5/12 00:00:02 90Nodes 24.c4 Qb4 25.Qh6+ Kg8 26.g5 Nh5 27.Rxh5 gxh5 28.Rxh5 Bf5 White is slightly better. +/= (0.48) Depth: 5/12 00:00:02 117Nodes 24.c4 Qb4 25.c5 dxc5 26.Qh6+ Kg8 27.g5 Nh5 28.Rxh5 gxh5 29.Rxh5 Bf5 White is slightly better. +/= (0.36) Depth: 6/12 00:00:02 318Nodes 24.c4 Qb4 25.c5 dxc5 26.Qh6+ Kg8 27.g5 Nh5 28.Rxh5 gxh5 29.Rxh5 Bf5 30.g6 White is slightly better. +/= (0.34) Depth: 6/13 00:00:02 360Nodes 24.c4 Qb4 25.c5 dxc5 26.Qh6+ Kg8 27.g5 Nh5 28.Rxh5 gxh5 29.Rxh5 Bf5 30.g6 Bxg6 White has an edge. = (0.28) Depth: 6/14 00:00:02 425Nodes 24.c4 a5 25.c3 a4 26.Nd2 h5 27.gxh5 Bf5+ 28.Ne4 Bxe4+ 29.fxe4 The position is equal. = (0.05) Depth: 7/14 00:00:03 1338Nodes 24.c4 a5 25.c3 a4 26.Nd2 h5 27.gxh5 Bf5+ 28.Ne4 Qc5 29.Qxc5 dxc5 30.hxg6 Nxe4 31.gxf7 The position is equal. = (0.03) Depth: 7/15 00:00:03 1822Nodes 24.c4 a5 25.c3 a4 26.Nd2 h5 27.gxh5 Bf5+ 28.Ka1 Nxh5 29.Ne4 Bxe4 30.fxe4 Qc5 31.Qg5 Qxc4 The position is equal. = (0.02) Depth: 7/16 00:00:03 1959Nodes 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qa3+ 28.Kb1 The position is equal. = (0.00) Depth: 7/17 00:00:03 2315Nodes 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qa3+ 28.Kb1 The position is equal. = (0.00) Depth: 7/18 00:00:03 2670Nodes 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qa3+ 28.Kb1 The position is equal. = (-0.01) Depth: 8/18 00:00:03 2779Nodes 24.Qh6+ Kg8 25.Ka1 Be6 26.Qg5 Kg7 27.Qh6+ Kg8 The position is equal. = (-0.08) Depth: 8/18 00:00:03 4064Nodes 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.02) Depth: 8/18 00:00:03 6407Nodes 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.06) Depth: 8/19 00:00:05 13kN 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.06) Depth: 8/20 00:00:05 13kN 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.06) Depth: 8/21 00:00:05 13kN 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.06) Depth: 8/22 00:00:05 14kN 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.06) Depth: 8/23 00:00:05 14kN 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.06) Depth: 8/24 00:00:05 14kN 24.Ka1 Be6 25.Qh6+ Kg8 26.Qg5 Kg7 27.Qh6+ The position is equal. = (-0.07) Depth: 9/24 00:00:06 17kN 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qc3 28.Ne4 Qa1+ 29.Kd2 Qd4+ 30.Kc1 Qa1+ The position is equal. = (-0.13) Depth: 9/24 00:00:08 24kN 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qc3 28.Ne4 Qa1+ 29.Kd2 Qd4+ 30.Kc1 Qa1+ The position is equal. = (-0.14) Depth: 9/25 00:00:08 25kN 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qc3 28.Ne4 Qa1+ 29.Kd2 Qd4+ 30.Kc1 Qa1+ The position is equal. = (-0.14) Depth: 9/26 00:00:08 25kN 24.c4 a5 25.Qh6+ Kg8 26.Nd2 Qb4+ 27.Kc1 Qc3 28.Ne4 Qa1+ 29.Kd2 Qd4+ 30.Kc1 Qa1+ The position is equal. = (-0.15) Depth: 9/27 00:00:08 26kN 24.Qh6+ Kg8 25.Ka1 Be6 26.Qg5 Kg7 27.Qh6+ Kg8 The position is equal. = (-0.15) Depth: 9/27 00:00:09 29kN 24.Qh6+ Kg8 25.Rh3 Be6 26.Qg5 Kg7 27.Qh6+ Kg8 The position is equal. = (-0.15) Depth: 10/27 00:00:09 29kN 24.Qh6+ Kg8 25.Rh3 Qa4 26.Rh4 Qc6 27.c4 Qxf3 28.Nd2 Qc3 29.g5 Qxd2 30.gxf6 Qxh6 31.Rxh6 Bc6 32.R1h3 Re6 33.Rxh7 g5 The position is equal. = (-0.17) Depth: 10/28 00:00:09 31kN 24.Qh6+ Kg8 25.Rh3 Qa4 26.Rh4 Qc6 27.c4 Qxf3 28.Nd2 Qc3 29.g5 Qxd2 30.gxf6 Qxh6 31.Rxh6 Bc6 32.R1h3 Re6 33.Rxh7 g5 The position is equal. = (-0.17) Depth: 10/29 00:00:09 32kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.f4 Qb4+ 30.Kc1 exf4 31.Rf1 Re5 32.Rhf2 Qa3+ 33.Kb1 f3 34.Nxf3 The position is equal. = (-0.17) Depth: 11/29 00:00:11 40kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rff2 Nc5 34.Qh3 exf4 35.Qh8+ Ke7 36.Qc3 The position is equal. = (-0.19) Depth: 11/30 00:00:12 48kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rff2 Nc5 34.Qh3 exf4 35.Qh8+ Ke7 36.Qc3 The position is equal. = (-0.20) Depth: 11/31 00:00:12 48kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rff2 Nc5 34.Qh3 exf4 35.Qh8+ Ke7 36.Qc3 The position is equal. = (-0.20) Depth: 11/32 00:00:12 48kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rff2 Nc5 34.Qh3 exf4 35.Qh8+ Ke7 36.Qc3 The position is equal. = (-0.20) Depth: 11/33 00:00:12 48kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd1 Nxf2+ 37.Rxf2 The position is equal. = (-0.21) Depth: 11/34 00:00:13 49kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd2 Qxc2+ 37.Ke3 The position is equal. = (-0.22) Depth: 11/35 00:00:13 51kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd2 Qxc2+ 37.Ke3 The position is equal. = (-0.22) Depth: 12/35 00:00:16 67kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd2 Qxc2+ 37.Ke3 Black has an edge. = (-0.23) Depth: 12/36 00:00:17 75kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd1 Nxf2+ 37.Rxf2 Qd4+ 38.Ke1 Black has an edge. = (-0.25) Depth: 13/37 00:00:20 88kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd1 Nxf2+ 37.Rxf2 Qd4+ 38.Ke1 Black has an edge. = (-0.27) Depth: 13/38 00:00:23 100kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Rf1 Qb4+ 30.Kc1 Qc3 31.f4 Ba4 32.Nb1 Qxc4 33.Rhf2 Ke7 34.f5 Nc5 35.fxg6 Nd3+ 36.Kd1 Nxf2+ 37.Rxf2 Qd4+ 38.Ke1 Black has an edge. = (-0.28) Depth: 13/38 00:00:28 125kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Re1 b5 30.Qh4 Ke7 31.cxb5 Bd5 32.c4 Qb4+ 33.Ka1 Qc3+ 34.Kb1 Be6 35.Qf2 Bf5+ 36.Ne4 Qxc4 37.Rc1 Qxb5+ 38.Ka1 Rb8 39.Qd2 Bxe4 40.fxe4 Qb4 41.Qxb4 Rxb4 42.Rc7 Rxe4 43.Rh8 Black has an edge. = (-0.27) Depth: 13/39 00:00:33 146kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Re1 b5 30.Qh4 Ke7 31.cxb5 Bd5 32.c4 Qb4+ 33.Ka1 Qc3+ 34.Kb1 Be6 35.Qf2 Bf5+ 36.Ne4 Qxc4 37.Rc1 Qxb5+ 38.Ka1 Rb8 39.Qd2 Bxe4 40.fxe4 Qb4 41.Qxb4 Rxb4 42.Rc7 Rxe4 43.Rh8 Black has an edge. = (-0.26) Depth: 13/40 00:00:33 146kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Re1 b5 30.Qh4 Ke7 31.cxb5 Bd5 32.c4 Qb4+ 33.Ka1 Qc3+ 34.Kb1 Be6 35.Qf2 Bf5+ 36.Ne4 Qxc4 37.Rc1 Qxb5+ 38.Ka1 Rb8 39.Qd2 Bxe4 40.fxe4 Qb4 41.Qxb4 Rxb4 42.Rc7 Rxe4 43.Rh8 Black has an edge. = (-0.26) Depth: 14/40 00:00:33 146kN 24.Qh6+ Kg8 25.c4 a5 26.Nd2 Bc6 27.g5 Nd7 28.Qxh7+ Kf8 29.Re1 b5 30.Qh4 Ke7 31.cxb5 Bd5 32.c4 Be6 33.Qf2 Qb4+ 34.Ka1 Qc3+ 35.Kb1 Bf5+ 36.Ne4 Qxc4 37.Rc1 Qxb5+ 38.Ka1 Rb8 39.Qd2 Bxe4 40.fxe4 Qb4 41.Qd5 Rb5 42.Qd3 Nc5 43.Qc4 Black has an edge. = (-0.29) Depth: 14/40 00:00:38 176kN
(, 02.07.2023)
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:04 pm
mwyoung wrote:
texium wrote:
mwyoung wrote:
texium wrote:
Fascinating game where CSTal crushes leela in a sicilian with the black pieces when white leaves book winning. Sacrifices expected. First move out of book is move 14 [pgn] [Event "CSTal vs Leela T1"] [Site "Lucas Chess R 2.08a"] [Date "2023.07.01"] [White "Lc0 v0.30.0"] [Black "CSTal-NN-Strong"] [Result "0-1"] [WhiteElo "3700"] [BlackElo "3700"] [ECO "B78"] [TimeControl "180+2"] [PlyCount "144"] [Opening "Sicilian Defence: Dragon Variation, Yugoslav Attack, Old Line"] [Termination "Mate"]
The game is fascinating but not for the reasons you are stating.
I took a few hours to study this game to understand what happened.
And on move 14 maybe out of your book, but the position is know theory. And White is not winning in this positions. The computers did not leave know theory until move 24 with the move 24.Qg5??.
And it is highly unlikely Lc0 played this game of chess without some kind of major problem. Lc0 played weaker then many FM human players in this game. And played some outright blunders. And I could not replicate the major errors of Lc0 0.30.0 even at 1 second a move.
In this game Lc0 played with a error rate of .49 centipawns per move average and this is not even GM strength. No modern computer would play this badly. Where Chess System Tal had a flawless error rate of .02 centipawns per move average.
I'm sorry but you're trying to use another engine ( presumably SF15.1 to analyze Leela which is a stronger engine. I used Leela 0.30 with the net T1, the strongest net 768x15 which is stronger than BT2 and BT3, and which data will soon be used, if it already isn't, to further make stockfish stronger. I'll reply with a screenshot of the lines Leela plays in that position. They left the provided opening book at move 14 though.
Error rate is the Error rate. And no chess engine today plays at a FM human level.
And why the losing move 24.Qg5. Leela does not want to play that move.
im so confused as to where you're getting this "FM Level" thing from. CPL means nothing if an engine blunders, and they do. SF, Dragon, and Leela have all been known to make large oversights, especially in blitz games. I could show you plenty of positions where SF and Leela just don't find the right move in a game, but that wouldn't make them weak engines.
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:05 pm
mwyoung wrote:
Here is my analysis
Enter New Game Line 0.0 4r3/pp1b1pkp/3p1np1/4p3/6P1/qNP1QP2/P1P4R/1K5R w - - 0 1
Analysis by Lc0 v0.30.0-rc1:
24.Qh6+ Kg8 White is slightly better. +/= (0.63) Depth: 1/2 00:00:02 2Nodes
what net are you using. i'm pointing out a weakness in the current leela net. Engine version is almost irrelevant since BT1, BT2, Bt3, and T1 all use it.
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:08 pm
texium wrote:
mwyoung wrote:
texium wrote:
mwyoung wrote:
texium wrote:
Fascinating game where CSTal crushes leela in a sicilian with the black pieces when white leaves book winning. Sacrifices expected. First move out of book is move 14 [pgn] [Event "CSTal vs Leela T1"] [Site "Lucas Chess R 2.08a"] [Date "2023.07.01"] [White "Lc0 v0.30.0"] [Black "CSTal-NN-Strong"] [Result "0-1"] [WhiteElo "3700"] [BlackElo "3700"] [ECO "B78"] [TimeControl "180+2"] [PlyCount "144"] [Opening "Sicilian Defence: Dragon Variation, Yugoslav Attack, Old Line"] [Termination "Mate"]
The game is fascinating but not for the reasons you are stating.
I took a few hours to study this game to understand what happened.
And on move 14 maybe out of your book, but the position is know theory. And White is not winning in this positions. The computers did not leave know theory until move 24 with the move 24.Qg5??.
And it is highly unlikely Lc0 played this game of chess without some kind of major problem. Lc0 played weaker then many FM human players in this game. And played some outright blunders. And I could not replicate the major errors of Lc0 0.30.0 even at 1 second a move.
In this game Lc0 played with a error rate of .49 centipawns per move average and this is not even GM strength. No modern computer would play this badly. Where Chess System Tal had a flawless error rate of .02 centipawns per move average.
I'm sorry but you're trying to use another engine ( presumably SF15.1 to analyze Leela which is a stronger engine. I used Leela 0.30 with the net T1, the strongest net 768x15 which is stronger than BT2 and BT3, and which data will soon be used, if it already isn't, to further make stockfish stronger. I'll reply with a screenshot of the lines Leela plays in that position. They left the provided opening book at move 14 though.
Error rate is the Error rate. And no chess engine today plays at a FM human level.
And why the losing move 24.Qg5. Leela does not want to play that move.
im so confused as to where you're getting this "FM Level" thing from. CPL means nothing if an engine blunders, and they do. SF, Dragon, and Leela have all been known to make large oversights, especially in blitz games. I could show you plenty of positions where SF and Leela just don't find the right move in a game, but that wouldn't make them weak engines.
Don't be confused. The question is why did Lc0 play the out right blunder. 24.Qg4??
And others in the game causing the bad error rating in the game.
Something is wrong with Lc0 as I have shown. Lc0 never plays the losing 24.Qg4??
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:11 pm
the funny thing is that by 1min leela has figured both Qh6 and Qg5 are both bad. But in blitz tc it thinks Qg5 is best. second pic at 15sec
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:12 pm
mwyoung. What leela net are you using, thats very important.
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:15 pm
what does your leela evaluate here : N2k1bnr/pp3ppp/8/4n3/4Ppbq/5N2/PPPPK1PP/R1BQ1B1R w - - 1 9
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:18 pm
texium wrote:
mwyoung. What leela net are you using, thats very important.
I guess your network is broken.
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:20 pm
mwyoung wrote:
texium wrote:
mwyoung. What leela net are you using, thats very important.
I guess your network is broken.
lol broken, i'm using that first net
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:21 pm
texium wrote:
mwyoung wrote:
texium wrote:
mwyoung. What leela net are you using, thats very important.
I guess your network is broken.
lol broken, i'm using that first net
LOL I am using the 2nd one.
2nd one tested better for testing
""In use cases with very low calculation time per move or slow hardware, a smaller network might be a better choice.""
Error rate calculation is amazing.
Last edited by mwyoung on Sun Jul 02, 2023 9:26 pm; edited 4 times in total
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 9:21 pm
Quote :
weird result, not only does dragon usually beat weaker engines, it also majorly underperformed here. Same with SF15.1 being below ethereal.
Which are the parameters you test ?
Time control Permanent brain Cores Hardware ?!
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 10:24 pm
mwyoung wrote:
texium wrote:
mwyoung wrote:
texium wrote:
mwyoung. What leela net are you using, thats very important.
I guess your network is broken.
lol broken, i'm using that first net
LOL I am using the 2nd one.
2nd one tested better for testing
""In use cases with very low calculation time per move or slow hardware, a smaller network might be a better choice.""
Error rate calculation is amazing.
what's your gpu
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 10:33 pm
Mclane wrote:
Quote :
weird result, not only does dragon usually beat weaker engines, it also majorly underperformed here. Same with SF15.1 being below ethereal.
Which are the parameters you test ?
Time control Permanent brain Cores Hardware ?!
just realized you used ponder on for your tests. i don't use that since leela can't. it has it in uci options but its fake, as noted on leela github i usually test 2|1, or 3|2. No ponder, 1T, i5-7500, GTX 1660 SUPER. i'm not a "tester" per se, but i have decent hardware and run personal tests. from a 2|1 small test, reb 16.2 got a win just because of an extremely imbalanced opening (using tcec 23 book)
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 11:08 pm
texium wrote:
mwyoung wrote:
texium wrote:
mwyoung wrote:
texium wrote:
mwyoung. What leela net are you using, thats very important.
I guess your network is broken.
lol broken, i'm using that first net
LOL I am using the 2nd one.
2nd one tested better for testing
""In use cases with very low calculation time per move or slow hardware, a smaller network might be a better choice.""
Error rate calculation is amazing.
what's your gpu
My hardware is stated, RTX 2080 TI, RTX 3070, AMD Threadripper 2950x, AMD 5900HX.
Your GTX GPU is the issue with Lc0 that I detected. You are using a huge net, and your GPU has no Tensor cores to run Lc0 at a fast speed.
Last edited by mwyoung on Sun Jul 02, 2023 11:29 pm; edited 1 time in total
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Chess System Tal 2 released Sun Jul 02, 2023 11:28 pm
texium wrote:
Mclane wrote:
Quote :
weird result, not only does dragon usually beat weaker engines, it also majorly underperformed here. Same with SF15.1 being below ethereal.
Which are the parameters you test ?
Time control Permanent brain Cores Hardware ?!
just realized you used ponder on for your tests. i don't use that since leela can't. it has it in uci options but its fake, as noted on leela github i usually test 2|1, or 3|2. No ponder, 1T, i5-7500, GTX 1660 SUPER. i'm not a "tester" per se, but i have decent hardware and run personal tests. from a 2|1 small test, reb 16.2 got a win just because of an extremely imbalanced opening (using tcec 23 book)
Ok but your hardware and testing conditions are far away from mine. My time control is LONGER. My hardware is FASTER. I use PONDER. My graphical cards are faster. If you sum up all this this is a quality issue in testing. I am sure others are better then me because i do not use super Hardware.
My point is that when you criticise things, you have to take care about hardware and parameters.
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Chess System Tal 2 released Mon Jul 03, 2023 2:31 pm
924 Partien gespielt / Turnier beendet Name des Turniers: Ryzen7 5700 Ort/ Land: LAPTOP-JTQQP8M7, Deutschland Spielstufe: Turnier 40/5 Hardware: AMD Ryzen 7 5700U with Radeon Graphics mit 4.096 MB Speicher
what time control is this, and does ethereal use more of your processing power than SF on 4T ponder? because sf is clearly struggling despite almost +100 on Ethereal in 40/15 TC. A lot of the results seem odd, as I mentioned. Leela is probably the most normal result since it is on dx12, t75 is used, and doesn't have a ponder setting.
texium
Posts : 125 Join date : 2022-07-19
Subject: Re: Chess System Tal 2 released Tue Jul 04, 2023 12:33 am
Chris, It may have been previously asked and answered but what was the reason for closed source code this time?
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: Chess System Tal 2 released Tue Jul 04, 2023 1:16 am
924 Partien gespielt / Turnier beendet Name des Turniers: Ryzen7 5700 Ort/ Land: LAPTOP-JTQQP8M7, Deutschland Spielstufe: Turnier 40/5 Hardware: AMD Ryzen 7 5700U with Radeon Graphics mit 4.096 MB Speicher
what time control is this, and does ethereal use more of your processing power than SF on 4T ponder? because sf is clearly struggling despite almost +100 on Ethereal in 40/15 TC. A lot of the results seem odd, as I mentioned. Leela is probably the most normal result since it is on dx12, t75 is used, and doesn't have a ponder setting.
The time comtrol is shown 40/5‘. Permanent brain is on. Stockfish , no matter if 15 or 16 cannot handle x in y . It sometimes has time forfeits because the arena gui is slow and stockfish has a very aggressive usage of time.
The on chip radeon card of this amd machine is not very strong for lc0. I have stronger graphical cards in the tower pcs. This 8 core is a tiny laptop.
mwyoung
Posts : 880 Join date : 2020-11-25 Location : USA
Subject: Re: Chess System Tal 2 released Tue Jul 04, 2023 1:45 am