In a recent game against a weak computer opponent on lichess, I (white) was in the following position:

[FEN "Q4k1r/p1Nbbppp/1pn1p3/8/5B2/5N2/P1K1nPPP/R6R b - - 0 1"]

The AI played the move Nb8 giving white a forced mate in 3, annotated by the lichess analysis board as a Nb8?! -- it looks like Bd8 is a better move by blocking the check with a supported bishop.

However, at least to my novice understanding, a move into a forced checkmate (let alone a relatively easy-to-see one in three moves) instead of a far better alternative should be a blunder, not a dubious move (which I understand as "questionable but possibly having merits"). Is there something I'm missing about the nature of this checkmate which makes Nb8 dubious but "possibly having merits", or is it a case of wrong automated annotation?

  • 8
    I don't know, but if I had Black's position I'd play an even worse move than Nb8: I'd play "resigns" which loses in 0 moves. – bof Feb 20 at 7:13
  • 1
    It's dubious because it shortens the path to mate. The game is lost in a few moves anyway. – TonyK Feb 20 at 20:49

A blunder is a game-changing move. Nb8 does not change the result of the game which is utterly and completely lost regardless of black's move. Earlier in the game black did blunder when it gave away its queen.

The reason for the "!" part of the "?!" (dubious move) notation in this case is because moving the knight opens up the line of sight of the d7 bishop.

  • 4
    ...which of course begs the question why not Nd8, with the same effect. (It would be interesting to know how the engine was dumbed down, to know why it played Nb8. Not that it matters in the position...) – Hauke Reddmann Feb 20 at 11:04

Computer annotations these days commonly use ?! to indicate moves which are noticeably worse than the best move, but not bad enough to warrant a ? or ??. There's not necessarily any merit to the move, as there would have been in the old meaning to the annotation.

The program I use to auto-annotate my games uses a threshold of 3 for ??, 1.5 for ?, and 0.5 for ?!. Other programs may use other thresholds, but the idea is the same.

The program I use will not use these annotations for the losing side once they're already losing by a sufficient amount. Apparently lichess will still use ?! but not ? or ?? annotations, in this situation. I believe this makes some sense, as the "best" move in a hopeless situation is not necessarily the move which postpones the mate for the longest. But the computer isn't that great at finding "merit" in objectively losing moves.

Can there be merit in a move that allows a forced mate? There wasn't in your case, but it can happen. I remember one game from my high school days where I was hopelessly losing (down a piece, and my opponent was about to promote) but I noticed I had a draw possibility if I could get my queen to a certain square. The catch was, it would take two moves to get there, and my opponent had a forced mate in 2 once I made the first queen move. I went for it anyway, he didn't notice the mate or my draw threat, and I stole a draw. Playing the "best" moves at that point would have been a certain loss.

I'm not quite sure why I still have the notation sheet for that game, but I do, so here's how that one ended:

[FEN "1Q6/1Npqkp1p/2p5/P3p3/2P1p3/1P5P/5PPK/8 w - - 0 36"] 

1.a6 Qf5?! 2.Qxc7+?? (2.Qd8+ Ke6 3.Nc5#) Kf8 3.a7?? Qf4+ 1/2-1/2
  • Thanks! Just to confirm since my understanding is still weak, was this a draw by perpetual check, or another mechanism? – nanofarad Feb 22 at 15:12
  • 1
    Well, kind of. Technically it was a draw by agreement when we both knew I had a perpetual. Also, technically perpetual check isn't in the rules; either you need to repeat the position 3 times, or go 50 moves with no captures and no pawn moves. But if you have a perpetual, one of those things will happen sooner or later. – D M Feb 22 at 23:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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