i definitely don't understand why a missed forced mate opportunity does not show up in the analysis --- it should be a blunder, or mistake, but instead - i really don't know if it even shows up. if it did, it was an inaccuracy, which simply is useless, afaict. the computer says: 4 inacuracies, 0 mistakes, 0 blunders. looks like a perfect game?!? NOPE!
on move 11 i castle; i missed a forced mate. i see it on the next move, thank gawd..... but, really, can someone explain why the computer analysis of this seems totally useless? thanks..............
I see that Stockfish commented, "Lost forced checkmate sequence. Bh7+ was best." on move 11.0-0-0, which it marks as "?!" - which means "dubious". This is strange as you say because it considers Black's 10.Re8 a blunder, and marks it "??".
Either way you'd think the program should have taken the value of a checkmate, some really large number, and subtracted the value of the move made to get some number x. Then depending on an internal table, used the value of x to decide on the annotation glyph and would then correspond to it being an "Inaccuracy, Mistake, or Blunder". Plus that value of x would seem to be needed to compute the Average centipawn loss. That might be the sticky point. Did the programmer not want to use a very large number in that computation and spoil and otherwise good centipawn loss because of a missed mate? And what if there was a mate in 3, but you mate in 4? Regardless, I agree that it should mark 11.0-0-0 as a blunder. On the other hand I think if you do a mate in 4, when there was a mate in 3, that it should not be a blunder. This complicates the coding, I'm sure.
When two moves are both completely winning, I don't think it makes sense to mark one as a blunder, even if the other is more efficient. For example, there's little difference between a move that is +8 and one that is +10 even though the difference between a move that is +0.5 and one that is +2.5 is huge.
@pawnmulch probably because although you lost the forced checkmate, your evaluation was still at like +10, so it probably didn't think there was a significant difference.
@zaneanderman makes sense........... i was just stuck on -------- wow, i missed a forced mate, that must mean i really really messed up...........
um, but still............ missing a forced mate should be a horrible blunder, in my book, esp if i never saw it and just missed the chance... i'm wondering if the little mistakes, blunders, inaccuracies are like not very accurate, and just broad brushes of how you played... when i play through the game using the analysis, and i miss the mate, it simply gives me a terrible number, which is pretty accurate, i'd say.
Inaccuracies/mistakes/blunders make more sense when the evaluation is closer to zero, let's say between +5/-5 (these numbers are invented but you get the idea). If the position is more lopsided than this, then usually most moves are winning, but the computer variations all lead to different positions (obviously) which look more or less crushing than each other, and the eval is quite volatile. In particular it often becomes hard to say one move is clearly better than another, when it might just be your prefered way to convert. So lichess won't include big 'swings' as mistakes/blunders when the position is totally winning for one side or the other already.
That said the system is quite accurate when you are within the 'normal range', i.e. the +/-5 mentioned earlier.
>10 is truncated as far as I know. The usual borders for errors (0.5, 1, 2) don’t apply.
its
good
right
-1 to +1 is 2 change which means the computer thinks it shifted x.and +11 to +13 is 2,the same x!the mathematical difference is same.
@ilampozhil : But the relative difference is not the same. The same logic is behind the advice you sometimes see to trade down when ahead in material. Your relative difference keeps getting larger. Example: Being behind one pawn and nothing else captured is not nearly as bad as being left in a lost King and Pawn against bare King.
