Quite a few articles have been written about questionable quality of analyses provided by the chess.com computer (strength 2500!). But the one I`d like to share is just unbloodybelievable. When I reached the position shown in the diagram I decided to make a move that I believed to lead to a winning endgame position and my opponent resigned later on.

Since I didn`t count moves before 34.Rxf7+ I was looking forward to the analysis to see whether I was right or not. I wasn`t surprised the move was marked as a blunder but the score was 0.2. I thought I had misjudged the position and won only because my opponent made a mistake. Since I saw none I followed the line suggested by the computer.

The final position is not drawish at all. The black king is too far away from the pawns to do anything. An average chess player would be able to beat a GM. Let us check some variations (the main line comes from my chess engine).

Thanks for reading. 

It's even worse than the example that I showed you recently ;)

Joey, would you call a black dog whiteish? The position is mate in 13 (I didn`t count it myself). The below position is mate in 15 - would call it drawish, too? 

if you plan on sacficing the exchange down, you should play ke2, as i believe their rook will be tied down to protecting the pawn, and if they move the king they will have to move it back to recapture the exchange sacrifice for the winning endgame..... computers are not always correct, most are not programmed to think like humans, but to think in terms of "probability".... also this is a good day, should computers become perfect, chess would become less interesting. I hate going to computers for advice; i prefer the lonesome suffering and yelling at my board "WHY CANT I FIGURE IT OUT!", to have my mom come in the room and yell "what are you doing!" (theres a lot of yelling in my house, but its all good natured :). )

It's conspicious that the chess.com computer is advertised with a 2500 rating. Current engines are much stronger even on smartphone-class hardware. The most reasonable explanation for this is that the analysis time must be extremely short (some fraction of a second per move).

So computer analysis should be used for most elementary blunderchecking only. Run all serious analysis offline on your own hardware.

@Bergami, thanks for your comment. You are right I should have played Ke2 first. I also admit my rook sacrifice was not necessary to win the game and it likely was not the best way to do so. But it was my way and it worked, didn`t it?  But I`ll remember your words. I kinda like simplifying endgames this way (and I sometimes misjudge the position and have to offer/accept draw) but I will give it a second thought next time.

I agree. But was my rook sacrifice a blunder? I don`t think so for I reached a won endgame. And I could give you more examples when a move is marked as a blunder and it is not - if you like read here (the second game, comment #13)

Rxf7 was certainly no blunder. It wasn't even a sacrifice but a liquidation.

The point is that you can be sure that it was no blunder since it was intended and the game proved you right. So you can safely ignore the engine output.

There pops up a thread on this issue almost every other day so the staff might consider to alot a few more CPU cycles to their analysis engine. There seems to be a demand for it.

I'm still waiting for an answer to my post on this very subject.

http://www.chess.com/forum/view/help-support/bizarre-computer-analysis

As it is advertised as an incentive to pay for Premium Membership, Chess.com should not ignore this issue.

I agree that it is not a blunder. in fact i exchange material down for won endgames all the time... However it must only be done if you are 500% certain it reaches a won position anyways... and be pumped up on endgame knowledge. I take pride out of know that i reach a winning endgame and there is literally no possible way my opponent can win (only draw if i mess up). the great thing about winning king pawn endgames.... THERE ARENT ANY TACTICS! nothing difficult to miss ^^.

Bergmami, I understand the need to simplify, I do it myself,  but it's often an illusion.

I totally disagree on your assertion that there aren't any tactics in pawn endgames: you find tactics in pawn endgames. The ideas are: checkmating, winning material, promoting a pawn. These endings can be very difficult with "mysterious" king's moves and I'd rather finish the game before if I can.

@Bergami & SuperCourgette (I like your new username, Bruno), no man can be really sure he will not make a mistake and draw or lose a "won" endgame. The game above is just my style and let me show you another really nice way how I simplify to reach a "won" endgame.

I agree the computer`s line 25...Qd3 is better than mine (I honestly didn`t even think of this move). I know that an engine or a better player than me would win the game in a more copybook way but I don`t mind my moves are marked as blunders, mistake, and inaccuracies as long as I know what I am doing.  

Just te explain the strange opening in the game (post #16). It is a thematic tournament - Hippopotamus Attack: