Chess.com's computer analysis

Flying Pig

Utrinque Paratus
Retired Moderator
Joined
Jan 24, 2009
Messages
15,647
Location
Perfidious Albion
I analysed a game today (I lost) and it came up with this:

Inaccuracies: 6 = 17.6% of moves
Mistakes: 4 = 11.8% of moves
Blunders: 11 = 32.4% of moves

What does all this mean? When you CA a game, what do the figures look like for you?
 
The computer plays through your game, giving each move a score. It also does its own search for that move. If the best move it finds gives a higher score than the move played in the game, then it adds a comment to the game that says inaccuracy, mistake, or blunder, and increments its total for that category.

The 3 categories are triggered by different amounts of gap between your move and the computer's. For example inaccuracy might be 0.1-0.25 points, mistake 0.25-0.5 points, and blunder >0.5 points. I'm not sure what the actual thresholds are, that was just a guess.

The stats are a lot less important than looking at the moves where the computer thought its move is better. I go through the analysis output and try to understand why the move actually played was inferior to the alternative.
 
I'm throughly unimpressed with chess.com's analysis engine & don't use it anymore.

http://www.chess.com/forum/view/community/the-chess-analysis-engines-blunders

http://www.chess.com/forum/view/general/the-chesscom-analysis-engine-is-a-jerk

From the second thread ("the chess.com analysis engine is a jerk" :D) :

esolom said:
The analyser said the following to me...

ALTERNATIVE - You slipped from a better position to an equal one. You should have played 7... e5

(+0.28) INACCURACY - Perhaps better was 11... exd5

ALTERNATIVE - You should probably give up chess.

BY THE WAY - I have been sleeping with your wife
That had me :lol:ing.
 
On topic the best "figures" I ever got from a chess.com analyzed game is from this one.

Inaccuracies: 0 = 0.0% of moves
Mistakes: 0 = 0.0% of moves
Blunders: 0 = 0.0% of moves

That said, the game was cut short prematurely when my opponent went on a mass resigning spree. My endgames are generally fairly full of inaccuracies & often mistakes. Personally I find the distinctions not very accurate. I've recieved a "??" for a move which chess.com rates as .10 worse than it's best idea. Panzar's chessmaster analysis is significantly superior, besides being overly critical of not-that-bad moves chess.com's engine often misses the best lines.
 
Back
Top Bottom