At what eval would most GMs resign?

Hi everyone, 

After many years away from the board, I have now rediscovered my fascination with the game, and I have a question. Here on Chess.com the computer analysis gives an eval for each and every move. Sometimes this eval shows +/- 6 in some of my games, and still, the player who at one point during the game was 6 points down finds a victory. I am certain that would not happen at GM level. But what is the maximum eval a GM level player would be able to tolerate without resigning? Is it -0.5, -1, -2, -3 or...? 

I'm looking forward to your replies. Thanks. 

- Stefan

As far as I know, many GMs would resign if even only a pawn down (so -1) in the endgame (not sure about the middlegame), but someone better qualified will hopefully comment.

It's not really about the evaluation being in numerical terms. The player looks at the position and if he is worse will look at whether there are 'practical chances' to create unsound but tricky counterplay, or if there are tricks and traps his opponent might miss in time pressure, or if it's possible to get to a drawn endgame despite being down a pawn or 2 for example. In GM games if they are a pawn down in a simple position they will resign.

Thank you for your replies. However, in some positions, the eval is in favor of the player who is behind on material. 

Here White is down 3 pawns. Still, the eval is +1.87. Would a GM resign in this position as Black (it is Black turns to move)? 

Here White is to move. The eval is -1.00. So would a GM make the move or simply resign? 

I will say again that a human being does not look at a position and think 'damn this is -1.87 I'm worse. Should I resign?'. A GM level player will though have a good understanding of positional factors and will not grab 3 pawns without realising the huge compensation this gives the other player.

It would never be in a GM game, at least at 1800 level, cos those people wont be that dumb to grab materials in opening without activating pieces. 

Some GMs like thinking very hard in a lot position. I don't know if the intention behind this is to irritate the opponent, but Hikaru definitely made it work against Caruana, even tweeting about it later.

Are you saying that Black didn't play well to reach this position? The computer analysis says Black is ahead by 1.00. 

I am aware that GMs do not think like that. I guess my question wasn't precise, so I'll try to rephrase it: 

If we look at all GM games ever played with one player resigning, what would the computer eval then be by the time of resignment? Would it be between -0.5 and -1? Between -1 and -3? Between -3 and 6? Between -6 and -12? Or...?

I guess the answer to your question is simply when, considering both their position and the strength of their opponent, they realise they are lost. And for the third and final time the numerical evaluation is irrellevant. 

depends on the clarity of the position. A gm will probably resign if they think they  somewhere between -1 and -2...but some positions are very 'messy' and still require accuracy from both sides, so they would probably play on.

1-1/2+ is considered winning position, (in most cases).  Please stop beating this dead horse.  

Everything is subject to special exceptions within the given position.  But everyone already knows that.  So again, please stop beating this dead horse.

If you have 1-2 pawn advantage, just --

1) avoid getting mated, 2) simplify down to a endgame and 3) win through pawn promotion.  Why is that so hard to conceptualize and believe?  Only on the internet?

Only on the internet do mindless topics become trending "HOT Topics."  Trending algorithms are designed to herd the sheep, make topics "hot," keep you eyeballs glued to the screen and to their ecosystem.  Neat trick.  It worked with this post too.    

I hope you are joking. I am just trying to learn. 

"Best" two introductory books on the market (both quick reads) are Paul Littlewood (Chess Tactics) and Jeremy Silman (Essential Chess Endings, Explained Move by Move).

Learn both books cold -- they will make you into a USCF @1700 player, very quickly.  It's a concise but essential body of knowledge.

Good Luck With It.  

In the game http://www.chessgames.com/perl/chessgame?gid=1008419, the analyzing GMs thought that white was winning when he resigned.  The main problem is that the player has to know the correct evaluation.  https://timkr.home.xs4all.nl/chess2/resigntxt.htm shows 35 games which the winning player resigned.  Most GMs will resign when they think they are lost, including no chance of a swindle.

When the players are lower rated, there is a greater chance of him/her/it making a blunder.  Above 1600, I would resign when down a rook, but below that I would hope for a mistake. Against anyone I would test their B+N mating knowledge.  Against players below 1000, I remove all my queen side pieces, including the queen, and still manage to win.

The quick answer is that the eval isn't a firm decision factor, but other elements must be considered.  However, when you know your opponent can win the game, you should resign.  For a beginner, I suggest to play every game to the end just for the learning possibilities.

A pawn down in an endgame may be an eval of 0.

If the extra pawn will cause a promotion, the eval will be closer to +50 (that's fifty, not five point zero) or it may even show mate.

This is also true in other positions, for example losing a rook during the first 10 moves, but having better development and a safer king may mean you're even winning in spite of the material deficit.

@thestefanhansen
Of course the eval itself is only part of the equation of resigning. Some positions that are objectively drawn are nearly impossible to draw in practice. Some positions that are objectively winning are nearly impossible to avoid a draw in practice.

Having said all that, GMs often resign surprisingly late. I'd say at least a whole piece (meaning non-pawn) worth of material and zero compensation plus the opponent has an easy plan to convert the advantage into a win. In other words both the computer and practical evaluation says you're hopelessly losing.

Nearly no one resigns when only a pawn or two down in the middlegame, even if they're objectively lost. There's too much play left.

Also if you've been putting energy into the game for the last 5 hours you're not going to give up lightly

You can find many GM games, even World Championship games, where the engine may say -3 or more in a position and the losing player has not resigned.  I notice it all the time.  You would not have to look at a lot of games to find an example.  Bottom line, they will resign when they are sure they have no practical chances.

Thanks a lot. I appreciate all your replies.  

Now we have an evaluation algorithm running constantly in Live Chess Games ??

What a BONEHEADED idea.

Watch any game Game in 3/0 between two player 2500+ and the evaluation function jumps all over the map (and at numbers above 1.5+) but the 2500+ players know it's all BS and they keep on playing, obviously.

Is this just another (nutty) layer of distraction from the V3 light show ??

Why should a screen for a chess game (which requires focus and concentration in order to follow GM play) read like a cable TV broadcast ??  Yikes !