Am wondering if it is possible for a move to be considered "good" by engines but is rejected by professional human players. One possible reason that I can think of is that the in order for this move to be successful, one has to play exactly right for the next 20 moves or so, which is easy for engines but difficult for human players.

I see three situations, where something like you describe could happen:

  1. wild tactical positions with open king, where moves have to be very precise and many humans might be too scared to enter such position
  2. Many composed studies feature surprising moves that could be rejected by humans. For instance studies in which you give up all your material except for say a bishop with which you mate somehow. In most positions you can't win with a bishop only, so lines leading to such positions would not be considered usually.
  3. some of the more tricky endgames feature very unnatural/inhuman moves that would be rejected by human players. One example are the computer lines after move 67 in game 6 of the last WC match Carlsen-Caruana. Some quotes regarding the winning lines suggested by computers:

Svidler:

Even when I saw what the machine suggested was winning there I honestly don't think it was humanly possible to find it on the 10 minutes that Fabi had, and it's incredibly weird how the position is conceptually lost after Kg6.

Kasparov:

The computer shows Black wins with 68..Bh4 here. But had Caruana played the incredible 69.Bd5 Ne2 70.Bf3 Ng1!! they would request metal detectors immediately! No human can willingly trap his own knight like that.

Svidler:

A move that a human is more or less incapable of making

Grischuk:

I would never even dream about Bh4

Carlsen:

I’m not going to disagree with the computer, I just don’t understand it!

  • 1
    Those quotes are great – Isac Dec 10 at 2:31

Yes, it's possible. See this article. The gist is, if for you're in a "draw and I win this tournament" situation, you would quite naturally play moves that might not be the best, but keeps the position simple and avoids tactical melees. You'd do this even if it reduces your chances to win the game. Conversely, if you're in a must-win situation, then you'll take chances and try to complicate the game even if they objectively weaken your position, because a draw is the same as a loss anyway. A computer on the other hand doesn't understand these things. It just makes the best moves.

Below is a quote from the article; see the article for the game and moves.

I showed the position to a number of players in Wijk aan Zee, and all gave me simple wins – for instance 31.Rxb7, 31.Rd7 or even 31.Rxf6. Remember, the first time control is looming and tournament victory is in grasp.

So what does our hero play? 31.Qa7?!! “Fritzy!” squealed Anand and went into uncontrollable fits of laughter when he saw this and the following moves (I filmed his mirth and included it in my multimedia report in ChessBase Magazine 69). He and the other players immediately recognised the “hand” of the computer.

If you switch on multiple-variation mode you will see that Fritz thinks 31.Qa7 is a tenth of a pawn better than the other alternatives. A computer program simply doesn't understand the difference between cast-iron moves that cannot fail, and a tight-rope walk on the edge of the precipice. I venture no human would undertake the latter course in the given tournament situation. Play through the continuation and judge for yourself.

  • 2
    I agree! In fact, this was how a player was found cheating in a xiangqi (chinese chess) tournament: Instead of keeping the situation simple while he has the advantage, the player chose an extremely complicated tactics combination and eventually won a "horse" (similar to a knight) after about 7 or 8 moves and thus won the game. The player did not even spend much time calculating the lines, which becomes very suspecious. Eventually he was found cheating with an engine. – Zuriel Dec 11 at 18:22

Well... absolutely possible and common. Magnus Carlsen himself said "I don't care" on his very positive position in the last game of the 2018 WCC match. Did he offer a draw in a position that a chess engine can very likely to win?

Engine might play stronger chess, but if we don't understand how to do it, it's pointless. Sometimes we don't always make the best objective move. Carlsen did it to his last game. Kramnik tried several unsound openings this year.

  • Thanks for yoru answer! I am also wondering if there exists a move which is recommended by engine but is rejected by most or all top human players. Not in a specific game, but the move is considered "unsound" or "unsafe" or "dubious" by most top players. – Zuriel Dec 8 at 14:09

Yes, it is very much possible and this is the very reason different openings exist!

Some examples:

  • 1 e4 as White is suggested by almost all engines. However, some GM's and WC feel that e4 is too committal.
  • Sicilian is almost never recommended by engines, whereas lot of GM's and WC's think that is the best way to equalize for Black. Engines usually suggest 1...e5 or 1...e6
  • Engine evaluation of KID may not always be understable to humans. See here: https://en.chessbase.com/post/review-mihail-marin-winning-with-the-kings-indian
  • Could you elaborate a bit on your answer? I don't quite see the connection to "moves rejected by humans". – user1583209 Dec 8 at 15:22
  • @user1583209 1 e4 is suggested by engines, but lot of GM's and WC's feel that e4 is too commital. – Leg Dec 8 at 15:53
  • Good answer I guess – Akash Roy Dec 8 at 16:05
  • Disagree with: "this is the very reason different openings exist!". Different openings existed before engines. – user1583209 Dec 8 at 16:32
  • @user1583209 I don't understand your comment. These openings continue to exist in the computer era means that humans make decisions differently than a computer and I have also give valid reasons. – Leg Dec 8 at 16:45

I would say definately yes. One example is the situation of a fortress. Even though one side is up in material the position is not winable. The engines can figure out that the fortress will not break in 10, 20 or whatever moves ahead but cannot figure out that it can never break! Then the evaluation function is obviously off and the suggested move is also off. Only a relatively strong human player can assert that it's a dead draw and reject the move and the line that follows.

  • How about the following one? Can engines usually figure that it is a draw? e4ec.org/images/immr/BlockedPawns.gif – Zuriel Dec 9 at 3:02
  • Also dont reject tablebases. Most of those fortress might be in tablebases – Isac Dec 10 at 2:32

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.