Can the world correspondence champion beat Stockfish

[JacquesRW]

Without computer help. no.

The question is with computer help and I strongly believe the reply is also no.

Therefore. you are suggesting that Stockfish can solve chess at 24 hours per move and all chess positions will be solved

If that is the case give Stockfish 24 hours per move and see if it can win with this position
Despite the clear superiority of engines, there ARE positions which chess engines don't (and possibly can't) understand that are quite comprehensible for human players. Typically these positions showcase the human ability to think creatively and formulate plans and understand long-term factors in the position.

An example is this position by Nobel Prize winner Roger Penrose, brother to GM Jonathan Penrose. Humans will easily see that White can do nothing, and Black will have no plausible attempts to win, but chess engines will assert that Black is easily winning due to the numerous extra black pieces on the board.

[pgn][Event "Computer chess game"]
[Date "2023.07.09"]
[Round "?"]
[White "Human"]
[Black "Stockfish-windows-x86-64-avx2"]
[Result "*"]
[BlackElo "3550"]
[Time "06:02:10"]
[WhiteElo "?"]
[TimeControl "0+30"]
[SetUp "1"]
[FEN "8/p7/kpP5/qrp1b3/rpP2b2/pP4b1/P3K3/8 b - - 0 1"]
[Termination "unterminated"]
[PlyCount "0"][/pgn]

The score given by Stockfish *from that position* is completely irrelevant, if it were to play out the game it would do so perfectly regardless, hence it understands the position just fine.

[Chessqueen]

Without computer help. no.

The question is with computer help and I strongly believe the reply is also no.

Therefore. you are suggesting that Stockfish can solve chess at 24 hours per move and all chess positions will be solved

If that is the case give Stockfish 24 hours per move and see if it can win with this position
Despite the clear superiority of engines, there ARE positions which chess engines don't (and possibly can't) understand that are quite comprehensible for human players. Typically these positions showcase the human ability to think creatively and formulate plans and understand long-term factors in the position.

An example is this position by Nobel Prize winner Roger Penrose, brother to GM Jonathan Penrose. Humans will easily see that White can do nothing, and Black will have no plausible attempts to win, but chess engines will assert that Black is easily winning due to the numerous extra black pieces on the board.

[pgn][Event "Computer chess game"]
[Date "2023.07.10"]
[Round "?"]
[White "Stockfish-windows-x86-64-avx2"]
[Black "Stockfish-windows-x86-64-avx2"]
[Result "1/2-1/2"]
[BlackElo "2000"]
[Time "20:53:36"]
[WhiteElo "2000"]
[TimeControl "0+5"]
[SetUp "1"]
[FEN "8/p7/kpP5/qrp1b3/rpP5/pP6/P3K3/8 b - - 0 1"]
[Termination "normal"]
[PlyCount "16"]
[WhiteType "program"]
[BlackType "program"]

1. ... Bf4 {(Be5-f4 Ke2-f3 Bf4-c1 Kf3-e4 Bc1-b2 Ke4-f4 Bb2-e5+ Kf4xe5)
0.00/82 4} 2. Kf3 {(Ke2-f3 Bf4-c1 Kf3-e4 Bc1-b2 Ke4-e3 Bb2-f6 Ke3-f4 Bf6-h4
Kf4-e4 Bh4-e1 Ke4-d5 Be1-g3 Kd5-e6 Bg3-f4 Ke6-d7 Bf4-h2 c6-c7 Bh2xc7
Kd7xc7) 0.00/89 5} Bc1 {(Bf4-c1 Kf3-e4 Bc1-b2 Ke4-f4 Bb2-e5+ Kf4xe5)
0.00/92 5} 3. Ke4 {(Kf3-e4 Bc1-b2 Ke4-f4 Bb2-e5+ Kf4xe5) 0.00/69 5} Bb2
{(Bc1-b2 Ke4-f4 Bb2-e5+ Kf4xe5) 0.00/68 5} 4. Kf4 {(Ke4-f4 Bb2-e5+ Kf4xe5)
0.00/66 5} Ba1 {(Bb2-a1 Kf4-e3 Ba1-e5 Ke3-e4 Be5-d6 Ke4-f3 Bd6-f4 Kf3xf4)
0.00/68 5} 5. Ke3 {(Kf4-e3 Ba1-e5 Ke3-e4 Be5-d6 Ke4-f3 Bd6-f4 Kf3xf4)
0.00/69 5} Bb2 {(Ba1-b2 Ke3-e4 Bb2-c3 Ke4-f4 Bc3-d4 Kf4-f3 Bd4-e5 Kf3-g4
Be5-b8 Kg4-f5 Bb8-g3 Kf5-e6 Bg3-c7 Ke6-d7 Bc7-f4 c6-c7 Bf4xc7 Kd7xc7)
0.00/61 5} 6. Ke4 {(Ke3-e4 Bb2-c3 Ke4-e3 Bc3-a1 Ke3-f4 Ba1-f6 Kf4-g4 Bf6-h4
Kg4xh4) 0.00/64 4} Ba1 {(Bb2-a1 Ke4-d5 Ba1-h8 Kd5-e6 Bh8-e5 Ke6xe5) 0.00/69
5} 7. Kd5 {(Ke4-d5 Ba1-h8 Kd5-e6 Bh8-e5 Ke6xe5) 0.00/66 5} Bb2 {(Ba1-b2
Kd5-e6 Bb2-e5 Ke6xe5) 0.00/69 5} 8. Ke6 {(Kd5-e6 Bb2-e5 Ke6xe5) 0.00/63 5}
Be5 {(Bb2-e5 Ke6xe5) 0.00/64 5} 9. Kxe5 {(Ke6xe5) 0.00/65 5 Stalemate}
1/2-1/2


1. ... Bf4 {(Be5-f4 Ke2-f3 Bf4-c1 Kf3-e4 Bc1-b2 Ke4-f4 Bb2-e5+ Kf4xe5)
0.00/82 4} 2. Kf3 {(Ke2-f3 Bf4-c1 Kf3-e4 Bc1-b2 Ke4-e3 Bb2-f6 Ke3-f4 Bf6-h4
Kf4-e4 Bh4-e1 Ke4-d5 Be1-g3 Kd5-e6 Bg3-f4 Ke6-d7 Bf4-h2 c6-c7 Bh2xc7
Kd7xc7) 0.00/89 5} *[/pgn]

The score given by Stockfish *from that position* is completely irrelevant, if it were to play out the game it would do so perfectly regardless, hence it understands the position just fine.

Sorry, but Stockfish does not understand the previous position with 3 Black Bishops and only if you remove 2 of those Bishops and let stockfish play out the game it would understand that it is a draw from the starting position, whereas any human player rated above 1000 would look at that position and after considering moves for Black it would offer a draw without making a single move knowing that it will not make any progress since the position is totally blocked

[jefk]

This topic has been discussed before, and it seems that some points have been overlooked by CQ in previous postings or responses. Regarding the artificial SF position, such positions don't typically arise from a normal computer-assisted correspondence game. Additionally, there are strong engines with MCTS analysis that can identify a draw in such positions, and some SF clones have been made with fortress detection, although they may need updating. Engines like Leela are MCTS and others like Shashchess have MCTS options.

As for the strongest theoretical correspondence chess, it is unclear what Deepmind has done with the latest Alphazero chess MCTS program, which played unconventionally with positional sacrifices. Perhaps if the latest version of this Alphazero were to run on very fast hardware, it might be able to beat SF in one odd game or so. However, for such a game, a fix would likely be found quickly, and such exceptions wouldn't occur in future games.

Nevertheless, this doesn't change the conclusion that at the highest level of analysis, chess is a draw, and this is not always apparent in correspondence chess. Even at lower levels, it becomes apparent that achieving a win is challenging due to the relatively high drawing margin in practical sense. While some slight advantages may arise, they don't tend to accumulate in the endgame due to Nnue and possibly tablebases. It is not so difficult to keep a draw with SF in these situations, but the chances of making a mistake can increase if the game is long and requires entering the one and only good move to maintain the draw.

For reference, here are some links to relevant ICCF events where these issues are being observed:

https://www.iccf.com/event?id=100104
and at lower level:
https://www.iccf.com/event?id=101454