No Mistakes, No Blunders... Yet Lost by Move 14 (and 93% accuracy) • page 1/2 • Lichess Feedback • lichess.org

Hey guys, I have some feedback regarding the game analysis.
Take a look at this game: Black loses quite early and decisively.

However, the analysis reports: 0 mistakes, 0 blunders, and 93% accuracy.
By move 14, the evaluation is already at -2.3.

Of course, it's possible to lose a game by playing many slightly inaccurate moves. But to my (beginner) eyes, this game doesn’t look like that.

https://lichess.org/hLRNXt6g/ Hey guys, I have some feedback regarding the game analysis. Take a look at this game: Black loses quite early and decisively. However, the analysis reports: 0 mistakes, 0 blunders, and 93% accuracy. By move 14, the evaluation is already at -2.3. Of course, it's possible to lose a game by playing many slightly inaccurate moves. But to my (beginner) eyes, this game doesn’t look like that.

#2

Shake234

#3

Yes, I know that I won as black :-)

But take a look at the stats for black:
0 mistakes, 0 blunders, and 93% accuracy

Yes, I know that I won as black :-) But take a look at the stats for black: 0 mistakes, 0 blunders, and 93% accuracy

SergioGlorias

#4

@Shake234 https://lichess.org/page/accuracy

#5

nadjarostowa

#6

Not sure what you are trying to say here...

You didn't make mistakes, and were winning... I think the black stats are pretty logical.

I could see a point for questioning the stats for White, though... it shows once more how questionable this whole thing is.

Not sure what you are trying to say here... You didn't make mistakes, and were winning... I think the black stats are pretty logical. I could see a point for questioning the stats for White, though... it shows once more how questionable this whole thing is.

PurpleInferno edited

#7

h3 was just barely not bad enough not to be considered an inaccuracy. As a human I'd still call it that, since it's obviously bad (white gets doubled pawns). But the computer neither knows nor cares about what's obvious.
Nc3 is similar, except now it's less obvious what's wrong with the move. The engine wants to play 9. Be3 Qb6 10. Nd2 Qxb2 11. Be2, sacrificing a pawn for inadequate compensation. It seems the problem with 9. Nc3 is your Nge7-Nf5 plan: after 9. Be3 Nge7 10. Bd3 white would be relatively fine, but he doesn't have time to play Bd3 after 9.Nc3, as illustrated by the fact that white lost a pawn in the game with 11. Bd3 (but the alternatives were just as bad - white's position really is quite dire already).
Bxd4, 17. Bxc6 and 20. Nb3 are called inaccuracies, but the evaluation difference is almost large enough for them to be mistakes. I'd say they're collectively about the same as two mistakes or one blunder.

And then you won a bunch of pawns almost by force and white was hopelessly lost.

You played well. Your opponent didn't.

7. h3 was just barely not bad enough not to be considered an inaccuracy. As a human I'd still call it that, since it's obviously bad (white gets doubled pawns). But the computer neither knows nor cares about what's obvious. 9. Nc3 is similar, except now it's less obvious what's wrong with the move. The engine wants to play 9. Be3 Qb6 10. Nd2 Qxb2 11. Be2, sacrificing a pawn for inadequate compensation. It seems the problem with 9. Nc3 is your Nge7-Nf5 plan: after 9. Be3 Nge7 10. Bd3 white would be relatively fine, but he doesn't have time to play Bd3 after 9.Nc3, as illustrated by the fact that white lost a pawn in the game with 11. Bd3 (but the alternatives were just as bad - white's position really is quite dire already). 12. Bxd4, 17. Bxc6 and 20. Nb3 are called inaccuracies, but the evaluation difference is almost large enough for them to be mistakes. I'd say they're collectively about the same as two mistakes or one blunder. And then you won a bunch of pawns almost by force and white was hopelessly lost. You played well. Your opponent didn't.

Shake234

#8

Sorry. Of course I was talking about the evaluation for white!

The point I am trying to make is this:

How is it possible, that black loses a game as fast and obvious and - at the same time - the Lichess Analysis doesn't find any mistakes and gives a very, very high accuracy.

I get that each move seems to be somehow "good enough to not considered a mistake" (from a mathematical point of view).
But no one should lose a game in 12 moves without any mistakes.

Sorry. Of course I was talking about the evaluation for white! The point I am trying to make is this: How is it possible, that black loses a game as fast and obvious and - at the same time - the Lichess Analysis doesn't find any mistakes and gives a very, very high accuracy. I get that each move seems to be somehow "good enough to not considered a mistake" (from a mathematical point of view). But no one should lose a game in 12 moves without any mistakes.

PurpleInferno

#9

White's average centipawn loss says he basically lost a pawn every 3 moves on average (well, not exactly because the evaluation isn't tied to material, but you get the idea). He just managed to do it by losing roughly a third of a pawn with most moves instead of making a bunch of good moves and a few bad ones.

And your play in the first 12 moves was basically perfect according to the computer: 4 average centipawn loss, 99% accuracy.

White's average centipawn loss says he basically lost a pawn every 3 moves on average (well, not exactly because the evaluation isn't tied to material, but you get the idea). He just managed to do it by losing roughly a third of a pawn with most moves instead of making a bunch of good moves and a few bad ones. And your play in the first 12 moves was basically perfect according to the computer: 4 average centipawn loss, 99% accuracy.

mkubecek

#10

The inaccuracy / mistake / blunder scale is very coarse and can be quite misleading so that I would suggest not to give the blunder/mistake/inaccuracy counts too much weight. The idea is that the evaluation is transformed into "winning chance" on the 0-1 scale (or 0-100%). A blunder is a move which reduces your winning chance by at 0.3 or more, a mistake by 0.2-0.3 and inaccuracy by 0.1-0.2.

One surprising consequence is that once your position is bad enough (winning chance below 0.3), whatever you play, it cannot possibly classify as a blunder. And another consequence is that with a bit of (bad) luck, it can take only 6 moves which do not classify even as inaccuracies to get from completely equal position to hopelessly lost. (You start at 0.5 and drop by less than 0.1 in each step.)

The inaccuracy / mistake / blunder scale is very coarse and can be quite misleading so that I would suggest not to give the blunder/mistake/inaccuracy counts too much weight. The idea is that the evaluation is transformed into "winning chance" on the 0-1 scale (or 0-100%). A blunder is a move which reduces your winning chance by at 0.3 or more, a mistake by 0.2-0.3 and inaccuracy by 0.1-0.2. One surprising consequence is that once your position is bad enough (winning chance below 0.3), whatever you play, it cannot possibly classify as a blunder. And another consequence is that with a bit of (bad) luck, it can take only 6 moves which do not classify even as inaccuracies to get from completely equal position to hopelessly lost. (You start at 0.5 and drop by less than 0.1 in each step.)