Computer chess is despised by grandmasters!

[matejst]

I can give practical examples of positions I would play despite being -2.00 and others I wouldn't play despite being +2.00. One has to know how computerchess translate to human chess, because it is quite different. That's the reason, among others, why I still look for "the perfect engine". E.g., most top engines value space much more than humans -- in the French, you can have a perfectly sound, but a bit cramped position, with a clear plan of play, despite an engines' -1,50, and, at the same time, a hopeless ending while the silicon monsters think it is a dead draw 0.00. I play most of the time against the Maia engines -- Maia 1500 is a tactically weak positional genius, e.g., and if one understand this, he can learn a lot from his games against it.

Just like all things in life, engines can be useful, or detrimental, and it's up to us to find the way to use them. And it is just like cars: there are some you can drive, some you can't: while I am not able to drive a F1, I can wisely enjoy my Renault. Imho, SF and its derivatives are useful only in certain cases -- most of the time it is much effective to use some "middle of the pack" engines, like Komodo 8, Orion 8, Wasp 5 (or 4,5), or, lately, Marvin and Zahak (Seer with the new net is also interesting). Most of these engines were developed by chess players and their evaluation -- HCE or the NN trained on this HCE -- were developed, at one point, from a human POV. They are not too fast, so the evaluation is based on positional elements we can assess too, not on positions achieved by the search 40 half-moves deep. Generally speaking, one can make a difference among engines written by good engineers, and engines written by chess lovers/players.

Then, OTB chess is a different beast completely: anybody who has played one season for his club understands well the tension, the fight in a human match, and also knows what the "price of moves" is in a game. The higher the price, the more difficult, the more passionating the game. Game 4, e.g., translated in a ending where the price of moves was very high -- I enjoyed it a lot.

Finally: grandmasters do no despise computers, engines, nor computerchess. They learn from their engines more than we do, and they understand computerchess better than most of us do.

[Ozymandias]

[GMs] understand computerchess better than most of us do.

But not as well as their amateur opposition in the Freestyle chess tournaments, as their withdrawal from the PAL/CSS tournaments made evident, over 15 years ago.

[Uri Blass]

[GMs] understand computerchess better than most of us do.

But not as well as their amateur opposition in the Freestyle chess tournaments, as their withdrawal from the PAL/CSS tournaments made evident, over 15 years ago.

There is no big money in freestyle tournaments so I guess no motivation for strong chess players to participate.

It is also not equal hardware tournament and I guess having a stronger computer is a significant advantage.

[mvanthoor]

If the best move on the board creates a position which is too complex for a human to calculate (= big chance for making blunders) or you simply don't know how to proceed (= big chance of choosing the wrong plan or going erratically between plans), then this best move as shown by the computer may not be the best move for the human.

I've often played games where I thought: "I think I could win this by sacrificing a minor piece for a pawn to break open the king-side and then go on a mating attack... but I could also just trade down so and so, and win the end-game."

Then I start to calculate because I "feel" there's a mate there, but fail to find it with certainty... and thus I choose to trade down, ending with a pawn up which is a passer on the 6th rank and I comfortably proceed to win the game.

Still the computer says that the minor piece sacrifice is the better move, because it leads to a forced mate in 9. As I couldn't see it, there's a chance I miss the "only move" somewhere in the sequence, the king escapes, and I end up two points down instead of one point up. I will then probably lose the end-game.

This is also what happens on the very highest levels, and this is the reason why even super-grandmasters often state that "no human would play this move." The "pointe" of the move is just too far in the future for a human to accurately judge, because you get confused. Even a super-grandmaster cannot see 30+ moves in the future to assess if a move ultimately works or not, but Stockfish can.

Therefore, when in doubt, play the move that gives you _some_ advantage for certain, instead of playing the move that _may_ give you an even bigger advantage (if you could actually see it and make sure of it).

[Eduard]

[GMs] understand computerchess better than most of us do.

But not as well as their amateur opposition in the Freestyle chess tournaments, as their withdrawal from the PAL/CSS tournaments made evident, over 15 years ago.

There is no big money in freestyle tournaments so I guess no motivation for strong chess players to participate.

It is also not equal hardware tournament and I guess having a stronger computer is a significant advantage.

Feestyle chess has an extremely high level. Such positions as in game 2 in Dubai can be given up on move 21.

I don't think fast hardware has any advantage. 12 cores are enough. I currently have access to 10 TB 7Men Syzygy. Maybe in January on all 17 TB 7men. I prefer that to hardware that is twice as fast.

As far as I know, grandmasters did not win the PAL tournaments on PlayChess. I don't think GMs have any better computer skills.

[Eduard]

the objectively best move is always the best move, especially when it comes to victory or defeat. If an GM and even the world champion can't find the best move, it doesn't mean that another player won't find the best move either. Who measures up to be able to judge that? If the best move wins objectively, then such a winning move must be recognized as a win (in the review and analysis with engines).

[dkappe]

the objectively best move is always the best move, especially when it comes to victory or defeat. If an GM and even the world champion can't find the best move, it doesn't mean that another player won't find the best move either. Who measures up to be able to judge that? If the best move wins objectively, then such a winning move must be recognized as a win (in the review and analysis with engines).

Except of course when there are multiple objectively best moves. We are a long way away of knowing what those are in most positions. What’s more useful is knowing the move that gives the most practical chances. If we stipulate that chess is a draw, then the objectively best move might be the one that allows the opponent the greatest chance of making a mistake. How do we determine what that is? We might need a MCTS engine to find those moves.

[Eduard]

Here is an analysis by GM Karsten Müller. He clearly states that Black is winning and he criticizes moves where the other GM is full of praise.

[Event "FIDE World Championship 2021"]
[White "Carlsen,Magnus"]
[Black "Nepomniachtchi,Ian"]
[Site "Dubai"]
[Round "2.1"]
[Annotator "Karsten Mueller"]
[Result "1/2-1/2"]
[Date "2021.11.27"]
[WhiteElo "2855"]
[BlackElo "2782"]
[PlyCount "116"]

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. g3 Be7 5. Bg2 0-0 6. 0-0 dxc4 7. Qc2 b5 8. Ne5 {Eine seltene Abweichung von der Hauptvarinate} (8. a4 {die aber einiges an
Gift hat.} )c6 9. a4 Nd5 (9... Qxd4 $2 10. Nxc6 $18 )10. Nc3 f6 11. Nf3 Qd7 12. e4 Nb4 13. Qe2 Nd3 {Monsterkraken werden in dieser Partie eine große
Rolle spielen. Der schwarze wird gar später eine Qualität gewinnen.} 14. e5 Bb7 15. exf6 Bxf6 16. Ne4 Na6 17. Ne5 $6 {Objektiv wohl noch spielbar, gibt
das der Partie aber eine ungünstige Richtung.} ({Sowohl} 17. Nxf6+ gxf6 18. Bh6 $14 )({als auch} 17. Be3 Nab4 18. Ne5 $14 {waren vorzuziehen.} )Bxe5 18. dxe5 Nac5 19. Nd6 $3 {Auch dieser Monsterkrake gehört in der Tat genau
hier hin.} Nb3 20. Rb1 $2 {Die Qualitätsopferidee ist zwar richtig, aber
diese Realisierung geht zu weit.} (20. Be3 Nxa1 21. Rxa1 {ist laut Computer
dynamisch ausgeglichen.} )Nbxc1 21. Rbxc1 Nxc1 22. Rxc1 Rab8 23. Rd1 Ba8 $2 (23... bxa4 24. Be4 g6 25. Qxc4 c5 $19 {gewinnt wegen der Fesselung in der
d-Linie.} )24. Be4 $2 {Das geht definitiv zu weit.} (24. Nxb5 Qb7 (24... Qe7 25. Nd6 )25. Be4 $15 {war angesagt.} )c3 $2 {Eine weitere Panikreaktion.
Beide Seite unterschätzen das Material.} ({Nach dem prophylaktischen} 24... g6 $1 {sollte Schwarz auf lange Sicht gewinnen, z.B.} 25. Nxb5 Qg7 26. Nd6 Qxe5 27. Nxc4 Qh5 $19 )25. Qc2 $6 {In Magnus Sinne nicht die passende Zugfolge.} (25. bxc3 bxa4 26. Qc2 $44 {war angesagt.} )g6 26. bxc3 $2 {Der Computer
bevorzugt an dieser Stelle} (26. Qxc3 bxa4 27. Rd2 $15 {um den Schaden zu
begrenzen.} )bxa4 $2 {Diese Stellungsöffnung am Damenflügel hilft
letztlich Weiß, der nun genug Kompensation hat.} (26... Qg7 {nebst Öffnung
am Königsflügel gewinnt für Schwarz, z.B.} 27. f4 g5 28. Qa2 gxf4 29. Qxe6+ Kh8 30. Nf5 Qg6 31. Qxg6 hxg6 32. Ne7 Rfd8 33. Nxg6+ Kg8 34. Ra1 Rd2 35. gxf4 bxa4 $19 )27. Qxa4 Rfd8 28. Ra1 c5 29. Qc4 ({Nicht jedoch} 29. Qxa7 $2 Qxa7 30. Rxa7 Rxd6 31. Bxa8 (31. exd6 Bxe4 32. f3 Bxf3 33. d7 Rd8 34. Rc7 Kf7 $19 )Rd2 $19 )Bxe4 30. Nxe4 Kh8 $1 {Der König geht prophylaktisch aus dem
Weg von Dxe6+ und Sf6+.} 31. Nd6 $1 {Der Springer geht auf seinen angestammten
Vorposten zurück.} ({Das gierige} 31. Nxc5 $2 Qd5 32. Qh4 Rf8 {bietet nur
Schwarz Chancen.} )Rb6 32. Qxc5 Rdb8 33. Kg2 a6 (33... Qc6+ {geht auch,
bietet Schwarz aber keine echten Gewinnchancen, weil nur noch Bauen am
Königsflügel verbleiben:} 34. Qxc6 Rxc6 35. Rxa7 Rxc3 36. Ne4 (36. Nf7+ $5 )Rcb3 37. Nf6 R3b7 $11 )34. Kh3 {Carlsens König ist hier bemerkenswert
sicher. Ein wichtiges Motiv. Denn Weiß möchte die Damen nicht tauschen,
damit die Türme ihre Stärke im Endspiel nicht zeigen können.} Rc6 (34... Qc6 35. Nf7+ Kg8 36. Qe7 Qb7 37. Nh6+ Kh8 38. Nf7+ $11 {war ein alternatives Ende
der Partie.} )35. Qd4 Kg8 36. c4 Qc7 37. Qg4 {Danach kann Schwarz den Druck
direkt rausnehmen.} ({Vielleicht war} 37. Ra2 {daher eine bessere praktische
Wahl.} )Rxd6 38. exd6 Qxd6 39. c5 Qxc5 40. Qxe6+ Kg7 41. Rxa6 Rf8 ({
Selbst} 41... Qf5+ $6 42. Qxf5 gxf5 43. Ra7+ Kg6 $11 {ist spielbar.} )42. f4 Qf5+ 43. Qxf5 Rxf5 ({Hier natürlich nicht mehr} 43... gxf5 $4 44. Kh4 Rf7 45. Kg5 $18 )44. Ra7+ Kg8 45. Kg4 Rb5 46. Re7 Ra5 47. Re5 Ra7 48. h4 (48. Kg5 {
wird mit} Kg7 $11 {beantwortet.} ({Selbst} 48... Ra2 $6 {ist spielbar:} 49. Rb5 Kg7 (49... Rxh2 $4 50. Kf6 $18 )50. Rb7+ Kg8 51. g4 Rf2 52. h4 Rf1 53. h5 gxh5 54. f5 Rg1 55. Kxh5 Rf1 $11 ))Kg7 49. h5 Kh6 ({Selbst} 49... gxh5+ $6 50. Kxh5 Ra1 {ist spielbar, aber diese Konstellation wird am Brett hin und wieder
verloren.} )50. Kh4 Ra1 51. g4 (51. hxg6 hxg6 52. g4 Rh1+ {ist ebenfalls total
remis.} )Rh1+ 52. Kg3 gxh5 53. Re6+ Kg7 54. g5 Rg1+ 55. Kf2 Ra1 56. Rh6 Ra4 57. Kf3 Ra3+ 58. Kf2 Ra4 1/2-1/2

[amanjpro]

Here is an analysis by GM Karsten Müller. He clearly states that Black is winning and he criticizes moves where the other GM is full of praise.

[Event "FIDE World Championship 2021"]
[White "Carlsen,Magnus"]
[Black "Nepomniachtchi,Ian"]
[Site "Dubai"]
[Round "2.1"]
[Annotator "Karsten Mueller"]
[Result "1/2-1/2"]
[Date "2021.11.27"]
[WhiteElo "2855"]
[BlackElo "2782"]
[PlyCount "116"]

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. g3 Be7 5. Bg2 0-0 6. 0-0 dxc4 7. Qc2 b5 8. Ne5 {Eine seltene Abweichung von der Hauptvarinate} (8. a4 {die aber einiges an
Gift hat.} )c6 9. a4 Nd5 (9... Qxd4 $2 10. Nxc6 $18 )10. Nc3 f6 11. Nf3 Qd7 12. e4 Nb4 13. Qe2 Nd3 {Monsterkraken werden in dieser Partie eine große
Rolle spielen. Der schwarze wird gar später eine Qualität gewinnen.} 14. e5 Bb7 15. exf6 Bxf6 16. Ne4 Na6 17. Ne5 $6 {Objektiv wohl noch spielbar, gibt
das der Partie aber eine ungünstige Richtung.} ({Sowohl} 17. Nxf6+ gxf6 18. Bh6 $14 )({als auch} 17. Be3 Nab4 18. Ne5 $14 {waren vorzuziehen.} )Bxe5 18. dxe5 Nac5 19. Nd6 $3 {Auch dieser Monsterkrake gehört in der Tat genau
hier hin.} Nb3 20. Rb1 $2 {Die Qualitätsopferidee ist zwar richtig, aber
diese Realisierung geht zu weit.} (20. Be3 Nxa1 21. Rxa1 {ist laut Computer
dynamisch ausgeglichen.} )Nbxc1 21. Rbxc1 Nxc1 22. Rxc1 Rab8 23. Rd1 Ba8 $2 (23... bxa4 24. Be4 g6 25. Qxc4 c5 $19 {gewinnt wegen der Fesselung in der
d-Linie.} )24. Be4 $2 {Das geht definitiv zu weit.} (24. Nxb5 Qb7 (24... Qe7 25. Nd6 )25. Be4 $15 {war angesagt.} )c3 $2 {Eine weitere Panikreaktion.
Beide Seite unterschätzen das Material.} ({Nach dem prophylaktischen} 24... g6 $1 {sollte Schwarz auf lange Sicht gewinnen, z.B.} 25. Nxb5 Qg7 26. Nd6 Qxe5 27. Nxc4 Qh5 $19 )25. Qc2 $6 {In Magnus Sinne nicht die passende Zugfolge.} (25. bxc3 bxa4 26. Qc2 $44 {war angesagt.} )g6 26. bxc3 $2 {Der Computer
bevorzugt an dieser Stelle} (26. Qxc3 bxa4 27. Rd2 $15 {um den Schaden zu
begrenzen.} )bxa4 $2 {Diese Stellungsöffnung am Damenflügel hilft
letztlich Weiß, der nun genug Kompensation hat.} (26... Qg7 {nebst Öffnung
am Königsflügel gewinnt für Schwarz, z.B.} 27. f4 g5 28. Qa2 gxf4 29. Qxe6+ Kh8 30. Nf5 Qg6 31. Qxg6 hxg6 32. Ne7 Rfd8 33. Nxg6+ Kg8 34. Ra1 Rd2 35. gxf4 bxa4 $19 )27. Qxa4 Rfd8 28. Ra1 c5 29. Qc4 ({Nicht jedoch} 29. Qxa7 $2 Qxa7 30. Rxa7 Rxd6 31. Bxa8 (31. exd6 Bxe4 32. f3 Bxf3 33. d7 Rd8 34. Rc7 Kf7 $19 )Rd2 $19 )Bxe4 30. Nxe4 Kh8 $1 {Der König geht prophylaktisch aus dem
Weg von Dxe6+ und Sf6+.} 31. Nd6 $1 {Der Springer geht auf seinen angestammten
Vorposten zurück.} ({Das gierige} 31. Nxc5 $2 Qd5 32. Qh4 Rf8 {bietet nur
Schwarz Chancen.} )Rb6 32. Qxc5 Rdb8 33. Kg2 a6 (33... Qc6+ {geht auch,
bietet Schwarz aber keine echten Gewinnchancen, weil nur noch Bauen am
Königsflügel verbleiben:} 34. Qxc6 Rxc6 35. Rxa7 Rxc3 36. Ne4 (36. Nf7+ $5 )Rcb3 37. Nf6 R3b7 $11 )34. Kh3 {Carlsens König ist hier bemerkenswert
sicher. Ein wichtiges Motiv. Denn Weiß möchte die Damen nicht tauschen,
damit die Türme ihre Stärke im Endspiel nicht zeigen können.} Rc6 (34... Qc6 35. Nf7+ Kg8 36. Qe7 Qb7 37. Nh6+ Kh8 38. Nf7+ $11 {war ein alternatives Ende
der Partie.} )35. Qd4 Kg8 36. c4 Qc7 37. Qg4 {Danach kann Schwarz den Druck
direkt rausnehmen.} ({Vielleicht war} 37. Ra2 {daher eine bessere praktische
Wahl.} )Rxd6 38. exd6 Qxd6 39. c5 Qxc5 40. Qxe6+ Kg7 41. Rxa6 Rf8 ({
Selbst} 41... Qf5+ $6 42. Qxf5 gxf5 43. Ra7+ Kg6 $11 {ist spielbar.} )42. f4 Qf5+ 43. Qxf5 Rxf5 ({Hier natürlich nicht mehr} 43... gxf5 $4 44. Kh4 Rf7 45. Kg5 $18 )44. Ra7+ Kg8 45. Kg4 Rb5 46. Re7 Ra5 47. Re5 Ra7 48. h4 (48. Kg5 {
wird mit} Kg7 $11 {beantwortet.} ({Selbst} 48... Ra2 $6 {ist spielbar:} 49. Rb5 Kg7 (49... Rxh2 $4 50. Kf6 $18 )50. Rb7+ Kg8 51. g4 Rf2 52. h4 Rf1 53. h5 gxh5 54. f5 Rg1 55. Kxh5 Rf1 $11 ))Kg7 49. h5 Kh6 ({Selbst} 49... gxh5+ $6 50. Kxh5 Ra1 {ist spielbar, aber diese Konstellation wird am Brett hin und wieder
verloren.} )50. Kh4 Ra1 51. g4 (51. hxg6 hxg6 52. g4 Rh1+ {ist ebenfalls total
remis.} )Rh1+ 52. Kg3 gxh5 53. Re6+ Kg7 54. g5 Rg1+ 55. Kf2 Ra1 56. Rh6 Ra4 57. Kf3 Ra3+ 58. Kf2 Ra4 1/2-1/2

I wonder why he didn't play in the candidates?

[amanjpro]

the objectively best move is always the best move, especially when it comes to victory or defeat. If an GM and even the world champion can't find the best move, it doesn't mean that another player won't find the best move either. Who measures up to be able to judge that? If the best move wins objectively, then such a winning move must be recognized as a win (in the review and analysis with engines).

That is not actually correct, and easy to disprove. We have many game pairs in TCEC between LC0 and Stockfish which ends up in a decisive game pair, eventhough both engines were playing the "best" move all the time. Which tells us, even with 104 Threads, 64GB of Hash, engines can play subotimial "bestmoves"

And also, many times I have seen SF's eval pointing at +2 and ended up drawing the game, again in TCEC