Odd Computer Analysis


I was studying the analysis of this game this afternoon when I was stumped by the computer analysis of my opponents move 27... Kxe7.  I realize that different programs will give a different analysis of positions, but this one was exceptionally unique.  For my 27th move I decided to exchange the queens.  I moved 27. Qxe7+.  The analysis told me that was a mistake, which I am okay with.  The problem is that my opponent completed the exchange by taking my queen (27... Kxe7) to which the analysis replied "(5.01) BLUNDER - Lucky you! Your opponent blundered! The best move was 27... Kc8".  The subsequent suggested lines result in black losing both rooks, the bishop, and two pawns in exchange for nothing. 

Has anyone seen a computer analysis like this one?  Was it trying to find a draw for black?  The "correct move" Kc8 it notes is Mat10, but I don't see how that was going to happen if white still had a queen.  Perhaps the computer was bored of explaining all of my little mistakes and just wanted the game over with faster?


That is very stupid! 



Yes the computer analysis needs to improve.  Do any chess.com staff have any explaination for that series of moves. 


Ya a lot of my analyses have crazy stuff like that...


I'm too tired to look at this one, but chess.com's computer analysis once rated me as having made a "mistake" (or maybe it was even a "blunder"?) in a wholly won endgame when I made a move that was technically incorrect but otherwise did not affect the outcome at all even right down to the number of moves to win.  The evaluation was -90 for me the whole time.

Yes, it was the technically wrong move.  Yes, if my opponent's king had been on a different square it would have been badbadbadbadbad.  But my opponent's king was not on a different square.

So I kind of found that amusing. (And have since learned to perform the pawn push correctly.)


that's odd...I too get very confusing lines....not as extreme as that example though.I think the computer analysis is faulty and needs improving.now I take the lines with a grain of salt and only accept the lines I agree with (especially the innacuriaces....I don't trust those at all)


That is the first analysis I have seen where there is an obvious error.  I've seen some inaccuracies and mistakes that I didn't completely understand, but that's more about my skill level than anything else.  Hopefully this is an isolated problem as I do prefer the look of the game analysis on chess.com to my Chessmaster 9000.  I suppose I'll just run them both for a little while.  Maybe Fritz is in my future.  Thanks for the comments everybody!


My guess is that the computer doing the analysis is set at a very fast level with a few seconds a move. the computer can also be over loaded depending on what else is going on reducing its quality.

I am curious what the program is and what the settings are for the analysis.

finally nice work questioning the computer and not trusting it blindly


I've also gotten what I thought was bad analysis on a number of occasions, like creating a discovered attack, and the computers suggests the 'correct' line is to ignore the hanging piece and do something else.

One analysis a few days ago was especially egregious, I created my own thread to see if someone could tell me why the computer says the best move is pushing a pawn, instead of forcing mate.  See if you can see somthing I don't.



TonyH, homernh, thank you both for your input.  homernh I did see your similar thread the other day, and I agree, Qg2 looks like an excellent move in that position.  I do not see anything better.  

The computer analysis feature certainly is convenient being just one click after the game and having it not take up any of my slow computer's resources.  It's helpful when I know I do not have much time to look over the game, or if it is a blitz game with blunders too numerous to count.  However, the staff here have always seemed to be very up front that the computer analysis is not very highly rated (~2500 I believe).  I do not know how long it looks at each move, but with the number of games submitted each day I'm sure it cannot be too long.  If we want a quick look over our games it is good enough, but for games that need a more thorough review there are plenty of options out there.

Feel free anyone reading this to comment about what program you like best to analyze games!

Thanks everyone!