Computer chess is despised by grandmasters!

[Eduard]

I read some GM analyzes on the current WCh match in Dubai. In some analyzes, human errors are played down! Whatever the world champion is playing is supposedly "fascinating". What does it matter if the engine says -1? "Here is a strong knight and a wonderful bishop on e4".

I studied a lot of computer games. I've been working on making books for many months, and with success! I get several thousand games a week that Stockfish plays with 64 cores. I invest a lot of hours to find new book variants.

For me there is no longer any doubt that the engines (especially Stockfish) plays significantly better than any human (I think 1000 Elo over all GMs). Stupid and fast, that was yesterday. The engines have learned an incredible amount since the neural network was around.

So if an engine in the middlegame shows an eval of +1 for one side, then that has not yet been won, but the direction is correct. As a person, you should then ask yourself: Why is Black +1 better? Instead of claiming that "White's position is humanly considered to play better". I think we can learn a lot from Engines now, also positionally. The engines have reached a new level of chess, and so we should trust the engine evals and let us show us why the engines rate one way and not another. In any case, I learn a lot from engines. You don't get better at chess by learning some new tactical tricks, but by learning a deeper understanding of positions. Today's engines show us this better than any human.

[Eduard]

Here is such an example of wonderful white human play, according to Grandmaster's comment:

[Event "Game 2: Magnus Carlsen - Ian Nepomniachtchi"]
[White "Magnus Carlsen"]
[Black "Ian Nepomniachtchi"]
[Site ""]
[Round ""]
[Annotator ""]
[Result "1/2-1/2"]
[Date "2021.11.27"]
[WhiteElo "2855"]
[BlackElo "2782"]
[PlyCount "116"]

1. d4 Nf6 2. c4 e6 { Nepomniachtchi is known to favour the Gruenfeld defence (with 2...g6). But Carlsen would certainly be well prepared for that, so it is not a big surprise that we see a different second move.
Carlsen now steers the game toward a Catalan, which Nepomniachtchi has already faced a few times in 2021. } 3. Nf3 d5 4. g3 Be7 5. Bg2 0-0 6. 0-0 dxc4 7. Qc2 b5 (7... a6 { is Black's most common move. } )8. Ne5 { A rare choice from Carlsen. } (8. a4 { is the most common move } )c6 9. a4 { White can win back the pawn immediately, but after } (9. Nxc6 Nxc6 10. Bxc6 Rb8 { Black is very comfortable. } )Nd5 { Blocking the diagonal of the Catalan bishop is the only reasonable option. } (9... Qxd4 10. Nxc6 Nxc6 11. Bxc6 Rb8 12. axb5 { is very bad for Black. The queen will soon be chased away from d4. } )10. Nc3 f6 11. Nf3 { One might consider keeping the diagonal open for the g2-bishop. But in fact } (11. Ng4 { would be a mistake, in view of } b4 { which forces White's knight back to a passive square, since } 12. Ne4 f5 { wins a piece } )Qd7 { A sensible move, supporting the pawn on b5 } (11... b4 12. Ne4 { transposes into a wild game Gelfand - Shirov, from 1992. } Ba6 13. Bh3 f5 14. Neg5 b3 15. Qd1 Rf6 16. e4 Nb4 17. exf5 exf5 18. Re1 Nc2 19. Nxh7 Kxh7 20. Ng5+ Kg6 21. Ne6 Qh8 22. Bxf5+ Kxf5 23. Qf3+ Kg6 24. Qe4+ Kf7 25. Ng5+ Kg8 26. Qxe7 Qh5 27. Ne6 Rxe6 28. Rxe6 Nxa1 29. Bh6 Qf7 30. Qg5 Kh7 31. Bxg7 Qxe6 32. Be5 Qg6 33. Qd8 Nd7 34. Qxd7+ Kg8 35. f4 c3 36. f5 cxb2 { Gelfand, Boris - Shirov, Alexei, 0-1, Paris Immopar rapid, 1992, https://lichess.org/STJ34wVf } )12. e4 Nb4 13. Qe2 Nd3 14. e5 { White has sacrificed a pawn, so it is natural that he tries to create threats while Black is not yet fully developed. } Bb7 15. exf6 Bxf6 16. Ne4 Na6 17. Ne5 (17. Nxf6+ gxf6 18. Bh6 Rf7 { is the computer's first choice, though by no means the only option. The position is very complex. } )Bxe5 (17... Qxd4 { doesn't work: } 18. Nxf6+ Rxf6 19. axb5 { is very bad for Black. } )18. dxe5 Nac5 { During the press conference, Carlsen admitted that he had overlooked this possibility. However, his exchange sacrifice is a good response to the situation. } 19. Nd6 Nb3 20. Rb1 { White had two ways to sacrifice the exchange. Carlsen (characteristically, perhaps) opts for the strategically 'clean' version, where both the Black knights get exchanged off. However, the 'messy' version was also interesting. } (20. Be3 { is the messy version, and well worth considering. } Nxa1 21. Rxa1 { leaves one dangerous knight on d3, but there are two pluses from White's point of view. The pair of bishops is an asset, and secondly a later b2-b3 may undermine the knight on d3. One possible continuation is } Nxe5 22. Bc5 Nd3 23. Nxb7 Qxb7 24. Bxf8 Rxf8 25. axb5 Qxb5 26. Qxe6+ Kh8 27. Rxa7 { with equality probably not far off. } )Nbxc1 { As an alternative, Anand mentioned a fascinating counter-sacrifice of the exchange. } (20... Rab8 21. Be3 c5 22. Nxb7 Rxb7 23. Bxb7 Qxb7 { when the powerful pair of knights would give Black good prospects. However, White could also consider declining the sacrifice. } )21. Rbxc1 Nxc1 22. Rxc1 { A fascinating situation has arisen. White has knight for rook and pawn, but what a beautiful knight on d6! The computer shows a preference for Black's position, but in practical play it is much easier to handle the White side. } Rab8 23. Rd1 { It is important to realise that White's compensation has a long-term character. There is no rush to recover the material. } (23. axb5 { is significantly weaker: } cxb5 24. Nxb7 Rxb7 25. Bxb7 Qxb7 { leaves White a clear pawn down } )Ba8 { It is very natural to tuck this bishop away and open the file for the b8-rook. But, ugly as it looks, the immediate } (23... bxa4 { was perhaps stronger. } 24. Be4 g6 25. Qxc4 c5 { is an important idea, using the pin on the d-file to exchange bishops. After, for example } 26. Rd2 Kh8 27. Qxc5 Bxe4 28. Nxe4 Qb7 { Black stands better, but there is a tough fight ahead. } )24. Be4 { A powerful move, which reveals a new dimension to White's compensation. There are serious chances of a mating attack on the kingside. Instead } (24. Nxb5 Qb7 { gives Black counterplay on the b-file } )c3 { At first sight, this move is hard to comprehend. Why give up a pawn? The purpose of this move is to open the b-file, to accelerate Black's counterplay in case White plays for mate on the kingside. Compare with the alternative: } (24... bxa4 25. Bxh7+ Kxh7 26. Qh5+ Kg8 27. Rd4 Rxb2 (27... Qe7 { is better: } 28. Rh4 Qxh4 29. Qxh4 { with wild complications } )28. Rh4 Rb1+ 29. Kg2 c5+ 30. Kh3 { and Black is lost, since he must give up the queen to avoid being mated. } )25. Qc2 (25. bxc3 bxa4 { and now, if White plays for mate, the purpose of 24...c3 becomes clear: } 26. Rd4 c5 27. Bxh7+ Kxh7 28. Rh4+ Kg8 29. Qh5 (29. Qc2 { is little better: } Rf5 { wins } )Rb1+ { mate! } )g6 { This weakens the seventh rank, and leaves a hole on f6 which the knight might later exploit. But } (25... h6 { has its own drawbacks, e.g. } 26. Qxc3 bxa4 27. Bc2 { and White is teeing up for Qc3-d3-h7 mate. } )26. bxc3 bxa4 (26... Qg7 27. f4 g5 { was interesting, and the computer favours Black. But who would willingly blow open the kingside, with all their pieces sitting passively on the back rank? } )27. Qxa4 Rfd8 { Defending the queen to allow the freeing move c6-c5. } 28. Ra1 c5 29. Qc4 Bxe4 30. Nxe4 (30. Qxe4 Rb2 { and White has nothing better than returning the queen to c4. Carlsen's recapture with the knight is more accurate, since the threat of Ne4-f6+ forces Black to make a king move, instead of this more active rook move. } )Kh8 31. Nd6 Rb6 32. Qxc5 Rdb8 33. Kg2 a6 { a patient move. Since sacrificing the exchange, Carlsen has collected two pawns, and at this point certainly has full compensation. But Nepomniachtchi has succeed in exchanging bishops and the rooks have gained in scope. The knight on d6 remains a powerful piece, but there is no obvious way for Carlsen to improve his position. } (33... Qc6+ { would be too hurried: } 34. Qxc6 Rxc6 35. Rxa7 Rxc3 36. Nf7+ Kg8 37. Ng5 { sees White collect a second pawn on the kingside. Black will not achieve a draw easily. } )34. Kh3 Rc6 35. Qd4 Kg8 36. c4 Qc7 37. Qg4 { This allows a liquidation, to a position where White has just a nominal advantage. That's an understandable decision since there was no obvious way to improve White's position. } Rxd6 38. exd6 Qxd6 39. c5 Qxc5 40. Qxe6+ Kg7 41. Rxa6 Rf8 (41... Qxf2 { would be a terrible blunder: } 42. Qe5+ Kh6 43. Qxb8 Qf1+ 44. Kh4 Qxa6 45. Qf8# )42. f4 Qf5+ 43. Qxf5 Rxf5 44. Ra7+ Kg8 { With all the pawns on the kingside, the extra pawn is not sufficient to win, as both players knew well. The final moves were played fairly quickly. } 45. Kg4 Rb5 46. Re7 Ra5 47. Re5 Ra7 48. h4 Kg7 49. h5 Kh6 (49... gxh5+ 50. Kxh5 { is still drawn, but Nepomniachtchi's move is safer. } )50. Kh4 Ra1 51. g4 Rh1+ 52. Kg3 gxh5 53. Re6+ Kg7 54. g5 Rg1+ 55. Kf2 Ra1 56. Rh6 Ra4 57. Kf3 (57. Kg3 Ra3+ 58. Kh4 Ra4 59. Rf6 h6 { is the simplest } 60. Kxh5 hxg5 61. Kxg5 { with a trivial draw } )Ra3+ 58. Kf2 Ra4 1/2-1/2

[dkappe]

If a line that an engine finds is too obscure or complex for someone to remember or understand over the board, it isn’t useful. I don’t think GM’s “despise” computer chess, they properly understand it for what it is: a tool to assist the hero of the story — the human chess player. The output of SF, Dragon and lc0 are only meaningful when understood and explained by strong human chess players.

The engine on your laptop isn’t a grandmaster any more than a moped in your garage is Usain Bolt.

[KLc]

I like mistakes. And humanity.

[Eduard]

If a line that an engine finds is too obscure or complex for someone to remember or understand over the board, it isn’t useful. I don’t think GM’s “despise” computer chess, they properly understand it for what it is: a tool to assist the hero of the story — the human chess player. The output of SF, Dragon and lc0 are only meaningful when understood and explained by strong human chess players.

The engine on your laptop isn’t a grandmaster any more than a moped in your garage is Usain Bolt.

In the example above, the grandmaster doesn't even try to explain why Black is better! He just writes that the computer rates it that way. A +1 is not a little! It's almost won! The fact that the white game is supposedly preferable here is, for me, contempt for objective facts.

[dkappe]

If a line that an engine finds is too obscure or complex for someone to remember or understand over the board, it isn’t useful. I don’t think GM’s “despise” computer chess, they properly understand it for what it is: a tool to assist the hero of the story — the human chess player. The output of SF, Dragon and lc0 are only meaningful when understood and explained by strong human chess players.

The engine on your laptop isn’t a grandmaster any more than a moped in your garage is Usain Bolt.

In the example above, the grandmaster doesn't even try to explain why Black is better! He just writes that the computer rates it that way. A +1 is not a little! It's almost won! The fact that the white game is supposedly preferable here is, for me, contempt for objective facts.

You’re getting too emotional. The engine has no emotions, any more than a hammer or a saw. If the +1 requires an engine’s calculating ability to convert into a win, then it’s not useful. It’s not really a +1 for flesh and blood chess players. That’s what’s important.

[Cornfed]

You…got some weird obsessive fetish, man. I’ll give you that.

GM Miguel Illescas is the GM, right?

[carldaman]

Strong humans can differentiate between good practical play and computer preferences, and will often choose lines of play that they can understand and handle over some nebulous engine evaluation, even at the expense of giving up chunks of centipawn value. We should accept it for what it is, but that requires a basic understanding of chess.

A weak player with a strong engine will not appreciate how grandmasters play and foolishly look down on them, despite not getting the why and how of either the GMs or the engines' moves.

[BrendanJNorman]

A weak player with a strong engine will not appreciate how grandmasters play and foolishly look down on them, despite not getting the why and how of either the GMs or the engines' moves.

This is essentially the gist of this thread and Chessqueen's thread about Nepo being "stupid for wasting time with a supercomputer".

How've you been Carl? Been a while man!

[Ozymandias]

If a line that an engine finds is too obscure or complex for someone to remember or understand over the board, it isn’t useful. I don’t think GM’s “despise” computer chess, they properly understand it for what it is: a tool to assist the hero of the story — the human chess player. The output of SF, Dragon and lc0 are only meaningful when understood and explained by strong human chess players.

The engine on your laptop isn’t a grandmaster any more than a moped in your garage is Usain Bolt.

In the example above, the grandmaster doesn't even try to explain why Black is better! He just writes that the computer rates it that way. A +1 is not a little! It's almost won! The fact that the white game is supposedly preferable here is, for me, contempt for objective facts.

You’re getting too emotional. The engine has no emotions

But you and me do, he does too. If you ask around, you'll hear most chess players saying that, yes, engines are stronger than humans, BUT... What follows, regardless of the argument, is emotional in essence. Let's say that engines go for a line which is too complicated for human play, you still have to accept that the stronger entity will be better qualified to evaluate the position. This doesn't happen when the human tiptoes around the engine's output. And why would he do that? either for emotional reasons or self-interest. If you accept on a regular basis, that you're basically translating what the engine says, you may find yourself out of a job, as there are translators around who aren't GMs.