Computer analysis - Difference between inaccuracy and mistake etc.

Sort:
Avatar of forked_again

In my game today my opponent threw away his queen.  Engine evaluation went from +5 to +16.  

I thought that difference between inaccuracy, mistake, and blunder, is the magnitude of how bad the move was, based on change in evaluation.  Obviously this is not the case.  So how are these defined?

Avatar of number-0

inaccuracy :meh you could do better

mistake: oh you shouldn't have done it. 

blunder: ARE YOU BLIND WHEN YOU PLAYED THAT MOVE !!!!!!

since it was +5 and it was already a lost i guess it was considered as a mistake or inaccuracy but if it was -5 to 0 thats a blunder by black

Avatar of forked_again

I can imagine a rule structure based on what you described number-0, but I'm wondering what the actual guidelines are.  And yeah, plus 5 is a pretty good advantage, but at the 1200 level anything can happen!  Lose your queen on top of that and then it it's definitely lost.  I can't imagine what the rule structure would be where giving away the queen and losing 11 evaluation points is not a blunder.

In general though I think you are right.  In the opening the computer will call a mistake for a decent opening move maybe .5 worse than the best.  Near the end of the game it doesn't call out small differences like that.

Avatar of number-0
forked_again wrote:

I can imagine a rule structure based on what you described number-0, but I'm wondering what the actual guidelines are.  And yeah, plus 5 is a pretty good advantage, but at the 1200 level anything can happen!  Lose your queen on top of that and then it it's definitely lost.  I can't imagine what the rule structure would be where giving away the queen and losing 11 evaluation points is not a blunder.

In general though I think you are right.  In the opening the computer will call a mistake for a decent opening move maybe .5 worse than the best.  Near the end of the game it doesn't call out small differences like that.

if its mate in 10 and you lost your queen by accident im pretty sure it would not be a blunder by the person losing the queen. those number thingy are actually useful for higher up games (GM FM IM etc)

Avatar of forked_again

Yeah but like I said it was a +5 evaluation not mate in 10.  +5 to +16.  That sounds like a blunder to me regardless of stage of the game.  Also, giving away your queen with no compensation is like the classic definition of blunder.  

Avatar of IcyAvaleigh
it is the difference between your move and the best move I think. idk the exact numbers but if the best move leads to 5+ and you make a move leading to a 3+ position or worse (2 points difference or more between your move and the best move), it is a blunder :)
Avatar of manekapa

https://support.chess.com/customer/portal/articles/1444907-in-the-computer-analysis-what-s-the-difference-between-inaccuracy-mistake-and-blunder-

Avatar of forked_again
IcyAvaleigh wrote:
it is the difference between your move and the best move I think. idk the exact numbers but if the best move leads to 5+ and you make a move leading to a 3+ position or worse (2 points difference or more between your move and the best move), it is a blunder :)

This is what I thought but that would make my example a blunder.  Going from plus 5 to plus 16 was called an inaccuracy.  

Avatar of forked_again

The explanation in this link also does not explain it.  It says that a 2 point difference from the best move is a blunder, unless it there is a massive lead alreay then 2 points doesn't matter.  It used the example of going from plus 30 to plus 28.  That is not remotely similar to going from plus five to plus 16 by blundering a queen.  It should be called a blunder, at least in my mind.  

Avatar of forked_again
 

 

Avatar of forked_again

So that's the situation I'm talking about.  Check the computer analysis.

Just FYI, I could not post this game using Chrome, as the post button was not there.  Had to switch to Edge

Avatar of forked_again

I should also point out when I played Bxd7  on move 23 the evaluation goes from +9.28 to +7, and the "deep" analysis calls that a good move, not even an inaccuracy.  

Avatar of forked_again

I guess I'm starting to think that deep analysis is pretty crappy analysis.

Avatar of forked_again

A quick analysis on Lucas chess using Komodo 9 shows I made 3 blunders (moves 17, 20, 23) and 2 mistakes (moves 24, 25).

This is compared to deep analysis which says 2 inaccuracies (move 4 and 25) and 1 mistake on move 17.  

Komodo also shows the queen blunder on move 25 by black as a blunder.

So obviously Lucas chess with Komodo (could also use stockfish which would probably be identical but just playing with Komodo lately) is a much better analysis and took less time to finish than the deep analysis on chess.com.