lichess.org
Donate

Incorrect blunders and mistakes in computer analysis

I think computer analysis should be equipped with some AI because it is currently artificially unintelligent.

If I make a move in a position which is valued at +15, but I make a move which is only +10 this is not a blunder by any stretch of the imagination, ot it is mate in 7, but I make a move which is mate in 15. This is not a blunder, no commentator of a high level game ( who has not access to an engine! ) would call such move a blunder. They will call it a brilliant winning move and praise the player for playing it safe instead of going into some double edged brilliancy which is difficult to see through.

A blunder is a move which turns a clearly winning position into a drawish position, or a drawish position into clearly a losing position. You should not throw blunders left and right in computer analysis, rather preserve it for moves which present a real turning point in the game.
Just one from my recent games:

http://en.lichess.org/ULa4bxL2/black#62

No human who has not access to an engine would call this move a mistake. How can you call this move a mistake? What moron would play Rb8 in this position ( which is allegedly not a mistake ) when they see that there is a clear way to win the pawn and have a very advantageous position?
Any "moron" who ever heard of open files in chess would play ... Rb8. That's one of the basic rules in chess, right ?
Not when there is an immediate rook move which wins material. Open files are general development rules for rooks. If you have nothing better to do you put your rook on an open file.

In this position the general concept of putting the rook on an open file - which in this particulare case does not win material in any apparent way - fails to explain why the concrete, immediate material winning move is a mistake. I'm sure there is some counter intuitive sequence by which putting the rook on the open file wins even more material, but in the given position this is by no means obvious. What is obvious is that white has a dangerous passed pawn in the stomach of black, which can be won in two moves.

I wonder how many players in a real game would play Rb8 here or when asked to annotate this game by just looking at the game ( not with an engine ) would give this move a question mark.
Well, the thing is, the move does win a pawn, but by changing bishops it transfer the game into a Rook only endgame, which are tricky even with a pawn advantage... if White do indeed avoid 35. Txf6, marked as the next error, the game is still playable

The suggested more Rb8, while less "human", keeps the Bishiops and the tension
What I really agree with is that white made a MISTAKE trading rooks.

Now, this is a mistake, which looks a mistake to a human. Why? Because it is evident that without rooks the position is simply lost for white, because black has a 2 - 1 pawn majority on the kingside and there is no way white can stop this majority and defend the center pawns at the same time so black will eventually eat up white's center pawns at will ( as it happened in the game ).
Your opponent missed the ...33. Bh8 move which would made things a bit harder for You in that position.
Hence the suggestion of taking open file with rook.
Here is the ultimate proof of me being correct:

http://en.lichess.org/EincZTLi#68

I made Stockfish play itself in the Arena chess GUI in demo mode after making my rook move. In the link the critical position is shown. Stockfish of course refuses to trade rooks, yet after playing the endgame against itself for almost 30 moves, Stockfish finally resigns with white. Please note that Arena plays quite slowly in demo mode, so these moves are analyzed very deeply, typically at least to depth 28. ( In the link there is a shallow lichess analysis of the rest of the game as well, which you can safely ignore, because the game was played in a much greater depth than that of lichess quick analysis. )
Main thing here is to see which move leads to faster win, and if difference is big enough to be marked as mistake instead of inaccuracy.
Plus, white could try avoiding exchange of bishops, and go for rook exchange instead. In that case game could be drawn, maybe.

This topic has been archived and can no longer be replied to.