Chess Players DO NOT Understand The EVALUATION BAR
08.10.2023
I had to make this video
145 Comments
Stockfish also doesn't do the best at evaluating moves in a "fight from behind" position since it's often better to play a move that is most likely to cause your opponent to respond inaccurately rather than the move that loses the slowest with best play by both sides.
For example, I had a (very bad) game some time ago where I blundered a Knight and a Rook, and my opponent was engaging in a King-hunt. Stockfish marked my move as a Blunder, with an eval of -M8 (Black has a forced win in 8 moves), preferring a move with an eval something like -17.8. However, the suggested move left me with no counterplay, while my move led to a Mate in 4 if Black bungled his mating combo (which he did, and I won).
a couple of corrections to some things you said in your video – since Stockfish version 15.1, the evaluation numbers no longer represent piece values such as +1.0 = a pawn advantage, or +3.0 = 3 pawns or a knight advantage – here is the new way to decode the Stockfish evaluation bar ….. +1.0 now means white has a 50/50 chance of winning (now this doesn't mean that black also has a 50/50 chance of winning, it doesn't, as part of the 50% of the time white would get a draw) — so over +1.0, means white has a better than 50/50 chance of winning, with perfect play by both sides. Less than +1.0 would mean that white is more likely to get a draw or a loss. The same logic can be applied to negative evaluations, so an eval of (minus) -1.0 would mean black has a 50/50 chance of winning, and so on.
You're talking nonsense, the recent version of Stockfish's evaluation function copies that of Leela chess zero I always monitor the development versions of Stockfish and with Stockfish 15.1 release , the contributors said that the eval bar of +1.00 is 50% chance of winning. You're just talking about Traditional Chess Engines.
2:00 wouldn't the eval bar be showing 0.00? this will be a draw because no pawns can move other than the pawn on c6 which if moved will just be taken by black. The only way for this to be in white's favor is if the black king takes it eye away from c7. The way for black to draw this is to simply move to c7 and then to any of these squares b8, c8, d8 back and forth. And because of the way the pawns are positioned there is no way for white's king to take any of black's pawn. If white moves pawn to c7 you simply take it and You are probably winning because you have a way to reach whites pawn but white will probably take opposition to stop you from advancing.
tldr: This position is draw because no one can advance any pawn realistically and no one can also take the other's pawn
According to engine attempting scholars mate is a mistake, while evidently against 200 elo player, thats the fastest mate and clearly better than playing the long game. So its basically an excellent move.
Finnally someone who actually understands that the eval bar. Btw i think a point you forgot, is that somrtimes even if you play thise moves that stockfish calculates tue eval bar will sometimes just equalise and you lose your advantage. This is because stockfish sometimes only calculates that specific position. Its more like if you dont play a move your winning by 1.4. If you play a move the complex position changes and your advantage reduces to 0.6
I suppose the “best” response in that situation would be an engine move, or even series of engine moves, that no person would ever think of in that type of game. If he exploited that line, there may be a lot of suspicion of cheating
Sometimes the eval bar is +999 because if you find the exact right moves you have a forced mate in 25 but then you make a single move and the bar is suddenly even because there was ONLY 1 MOVE that won, everything else was drawing.
Perhaps rather than blunder we should say a mistake. If a position is objectively winning and the player doesn't play the winning line that's a mistake, even if it's incredibly difficult for a human to spot.
So whats the point of this video? The move of the grandmaster is not bad because because they are not capable of playing the best moves all the time? The evaluation bar is an objectively evaluation of the position and if a move is objectively bad then it is bad! We all understand the evaluation bar pretty much.
I hope people don’t make the mistake of thinking that it’s ok to disregard the evaluation bar, because this video is implying you should essentially assume your opponent will blunder, you should approach every game as though you are playing stockfish or any other super bot
so the point is that humans cant find the counter move, so it was not a mistake? disagree. the video could have argued instead that the eval bar is only an estimate, only mate in X moves is when the status is known
Putting aside that nowadays blunders are defined by engine eval regardless of how "findable" it is for a human, Rg8 is still a blunder from a human perspective. The whole point of b4 the move prior is to distract the Queen from f3 to play Nf3. Now the line gets crazy complicated after that and it's unreasonable to expect a mortal player to calculate it out to the end as winning (although I think if played out an SGM should be able to navigate it to victory) but you can't play b4 and then not go into that line. Ian lost because he didn't trust whatever evaluation he made a move earlier that caused him to play b4 and it cost him a world championship.
ussually theese mooves aren't blunders or even innacuricy's, just classified as "good" or "exelent" moves, which loose some dvantage because there not the best
Except it is the world championship and such "crazy" lines have been found by previous world champions and candidates before, it is objectively a blunder because it loses advantage. It is expected of players of those calibers to find those lines to make history, instead what we got was an f5 blunder by Ian losing the game after being objectively winning for like 6 moves. Ian chose the complicated line in the game you are showing and couldnt calculate enough, still think Magnus would have found it
The only reasonable definition of blunder is that of a move that makes the side worse in terms of the game's final result: if your move turns your position from drawish to losing, or from winning to drawish or losing (all by force) that move is a blunder. With this definition of a blunder, every move which is not a blunder is a "good move" because it doesn't worsen your final result. Moreover, doesn't exist a move that improves your final result from drawish to winning or from losing to drawish or winning; in fact, if your position is losing (by force) that means that there exists at least one forced sequence of moves that leads to your defeat. If your opponent chooses the correct path you can't do anything about that. The same goes for a drawish position: if a position is drawish it means that neither side can force a win, that is, a checkmate within all the rules of chess. At most, we can say that, between two good moves, one is "better" than the other if the first leads to a forced win through a shorter sequence of moves, but in the end what really matters is the final result: if you win you gain 1 point no matter of many moves you have played to obtain the victory. This makes it very complicated to correctly define a "brilliant" move, an "excellent" move and so on. They are all good moves in the sense that they are not blunders. On the other side, an inaccuracy, a mistake and a blunder are really only different terms to describe a wrong move, that is, a move which worsens your final result.
This is obvious, the thing is chat won’t go “omg, he didn’t find the 10 moves winning combination and gave away his advantage”, it’s obviously easier to say “he blundered”
A BLUNDER is a terribly bad move. That word should not be used for all moves that (according to stockfish) goes from a winning position to a losing one. Because in human terms any move that looks good for 5-10 moves into the future is still considered good (more moves required in case of simple positions).
Even so. It should be considered that the stockfish analysis has an excellent success rate. And stockfish clearly tells us that in his analysis a position goes from winning to losing(or draw)
I think I see your point, but you just didnt explain it that well, it was indeed a blunder, the computer says it's one so it is one, I think your opinion is that people screaming "blunder!" 23 ahead is completely irrelevant when in some situation gayfish sometimes finds lines that are almost impossible for a human to find, people should analyse the "blunder" and try to play the line themselves before acting like it's the end for white
Chess has 100^120 possible moves not 100^40. To clarify, in our universe, there are about 100^80 atoms. There are more more possible moves than atoms in our universe.
You didn't explain anything we didn't already know.
It is a blunder if it wipes away a couple pawns' worth of value from your position, and the value isn't far from zero. What you stated is just false. It's irrelevant that we're human and have difficulty finding certain computer lines.
Stockfish also doesn't do the best at evaluating moves in a "fight from behind" position since it's often better to play a move that is most likely to cause your opponent to respond inaccurately rather than the move that loses the slowest with best play by both sides.
For example, I had a (very bad) game some time ago where I blundered a Knight and a Rook, and my opponent was engaging in a King-hunt. Stockfish marked my move as a Blunder, with an eval of -M8 (Black has a forced win in 8 moves), preferring a move with an eval something like -17.8. However, the suggested move left me with no counterplay, while my move led to a Mate in 4 if Black bungled his mating combo (which he did, and I won).
Stockfish actually considers bishops & knights to be over 4 points and rook like 6
a couple of corrections to some things you said in your video – since Stockfish version 15.1, the evaluation numbers no longer represent piece values such as +1.0 = a pawn advantage, or +3.0 = 3 pawns or a knight advantage – here is the new way to decode the Stockfish evaluation bar ….. +1.0 now means white has a 50/50 chance of winning (now this doesn't mean that black also has a 50/50 chance of winning, it doesn't, as part of the 50% of the time white would get a draw) — so over +1.0, means white has a better than 50/50 chance of winning, with perfect play by both sides. Less than +1.0 would mean that white is more likely to get a draw or a loss. The same logic can be applied to negative evaluations, so an eval of (minus) -1.0 would mean black has a 50/50 chance of winning, and so on.
That's not how advantage works, stockfish uses win percentage, not material
You're talking nonsense, the recent version of Stockfish's evaluation function copies that of Leela chess zero I always monitor the development versions of Stockfish and with Stockfish 15.1 release , the contributors said that the eval bar of +1.00 is 50% chance of winning.
You're just talking about Traditional Chess Engines.
2:00 wouldn't the eval bar be showing 0.00? this will be a draw because no pawns can move other than the pawn on c6 which if moved will just be taken by black. The only way for this to be in white's favor is if the black king takes it eye away from c7. The way for black to draw this is to simply move to c7 and then to any of these squares b8, c8, d8 back and forth. And because of the way the pawns are positioned there is no way for white's king to take any of black's pawn. If white moves pawn to c7 you simply take it and You are probably winning because you have a way to reach whites pawn but white will probably take opposition to stop you from advancing.
tldr: This position is draw because no one can advance any pawn realistically and no one can also take the other's pawn
According to engine attempting scholars mate is a mistake, while evidently against 200 elo player, thats the fastest mate and clearly better than playing the long game. So its basically an excellent move.
"Ee-an" its "y-an"
This entire video can be summed up to: they may have not been blunders, just inaccuracies.
Finnally someone who actually understands that the eval bar.
Btw i think a point you forgot, is that somrtimes even if you play thise moves that stockfish calculates tue eval bar will sometimes just equalise and you lose your advantage. This is because stockfish sometimes only calculates that specific position. Its more like if you dont play a move your winning by 1.4. If you play a move the complex position changes and your advantage reduces to 0.6
I suppose the “best” response in that situation would be an engine move, or even series of engine moves, that no person would ever think of in that type of game. If he exploited that line, there may be a lot of suspicion of cheating
Sometimes the eval bar is +999 because if you find the exact right moves you have a forced mate in 25 but then you make a single move and the bar is suddenly even because there was ONLY 1 MOVE that won, everything else was drawing.
i think that u dont understand the eval bar ngl
Centipawn is no longer pegged to a pawn.
Great video !
You made a whole video trying to say that the bar does not tell you who is winning and then proceed to conclude that it actually does LMAO
Perhaps rather than blunder we should say a mistake. If a position is objectively winning and the player doesn't play the winning line that's a mistake, even if it's incredibly difficult for a human to spot.
Paying attention to engines so much is one of the problems with modern chess.
LMFAO the subs call Ding "Dangly Wren"
Chess is a game of perfect information. That means there is always at least one objectively best move. That is what the evaluation bar shows.
0:00 That is certainly a way to pronounce their names. (God, listening to that hurts.)
We're only human after all.
That 800 elo player who predicted the blunder ,XD
Cuz he has mastered the art of blundering
So whats the point of this video? The move of the grandmaster is not bad because because they are not capable of playing the best moves all the time? The evaluation bar is an objectively evaluation of the position and if a move is objectively bad then it is bad! We all understand the evaluation bar pretty much.
I hope people don’t make the mistake of thinking that it’s ok to disregard the evaluation bar, because this video is implying you should essentially assume your opponent will blunder, you should approach every game as though you are playing stockfish or any other super bot
so the point is that humans cant find the counter move, so it was not a mistake? disagree. the video could have argued instead that the eval bar is only an estimate, only mate in X moves is when the status is known
did people not know this?
they should make a stunkfish which only looks like 5 moves ahead, and then evaluates.
Putting aside that nowadays blunders are defined by engine eval regardless of how "findable" it is for a human, Rg8 is still a blunder from a human perspective. The whole point of b4 the move prior is to distract the Queen from f3 to play Nf3. Now the line gets crazy complicated after that and it's unreasonable to expect a mortal player to calculate it out to the end as winning (although I think if played out an SGM should be able to navigate it to victory) but you can't play b4 and then not go into that line. Ian lost because he didn't trust whatever evaluation he made a move earlier that caused him to play b4 and it cost him a world championship.
Bro really took twitch chat serious 💀
ussually theese mooves aren't blunders or even innacuricy's, just classified as "good" or "exelent" moves, which loose some dvantage because there not the best
You pronounce Ian in 1 Syllable. Not 2. His name is Ян, Not Иан
Except it is the world championship and such "crazy" lines have been found by previous world champions and candidates before, it is objectively a blunder because it loses advantage. It is expected of players of those calibers to find those lines to make history, instead what we got was an f5 blunder by Ian losing the game after being objectively winning for like 6 moves. Ian chose the complicated line in the game you are showing and couldnt calculate enough, still think Magnus would have found it
the best move is not always the best move
Please don’t say ean.
The only reasonable definition of blunder is that of a move that makes the side worse in terms of the game's final result: if your move turns your position from drawish to losing, or from winning to drawish or losing (all by force) that move is a blunder. With this definition of a blunder, every move which is not a blunder is a "good move" because it doesn't worsen your final result. Moreover, doesn't exist a move that improves your final result from drawish to winning or from losing to drawish or winning; in fact, if your position is losing (by force) that means that there exists at least one forced sequence of moves that leads to your defeat. If your opponent chooses the correct path you can't do anything about that. The same goes for a drawish position: if a position is drawish it means that neither side can force a win, that is, a checkmate within all the rules of chess. At most, we can say that, between two good moves, one is "better" than the other if the first leads to a forced win through a shorter sequence of moves, but in the end what really matters is the final result: if you win you gain 1 point no matter of many moves you have played to obtain the victory.
This makes it very complicated to correctly define a "brilliant" move, an "excellent" move and so on. They are all good moves in the sense that they are not blunders. On the other side, an inaccuracy, a mistake and a blunder are really only different terms to describe a wrong move, that is, a move which worsens your final result.
This is obvious, the thing is chat won’t go “omg, he didn’t find the 10 moves winning combination and gave away his advantage”, it’s obviously easier to say “he blundered”
A BLUNDER is a terribly bad move. That word should not be used for all moves that (according to stockfish) goes from a winning position to a losing one. Because in human terms any move that looks good for 5-10 moves into the future is still considered good (more moves required in case of simple positions).
Even so. It should be considered that the stockfish analysis has an excellent success rate. And stockfish clearly tells us that in his analysis a position goes from winning to losing(or draw)
Let me make fun of an 800 rated chatter so people won't notice how incredibly fucking stupid my argument's about to be
it usually needs some time to have balanced
'It's not a blunder, it's a mistake'
I think I see your point, but you just didnt explain it that well, it was indeed a blunder, the computer says it's one so it is one, I think your opinion is that people screaming "blunder!" 23 ahead is completely irrelevant when in some situation gayfish sometimes finds lines that are almost impossible for a human to find, people should analyse the "blunder" and try to play the line themselves before acting like it's the end for white
0:17 This is a bad example because Ian really lost all his advantage after this move
Chess has 100^120 possible moves not 100^40. To clarify, in our universe, there are about 100^80 atoms. There are more more possible moves than atoms in our universe.
You didn't explain anything we didn't already know.
It is a blunder if it wipes away a couple pawns' worth of value from your position, and the value isn't far from zero. What you stated is just false. It's irrelevant that we're human and have difficulty finding certain computer lines.
In the thumnail, Black King is moving down to threaten White's Knight. He's not in Checkmate yet, but he will likely lose his queen for a knight.
Not you tho
-Mayhem Chess
Bro what a useless video, everyone knows that, the people in chat just expect them to play better (which is dumb but happens with any e-sport).
Thumbnail is clickbait! You dont need to watch a single second of the video to know that lol
Good video (just note that Ian is pronounced Yaan, or you might offend some opeople)