I havent checked to see if this is already an ongoing topic of discussion, but I have a small bone to pick with chess.com, and I was curious to know if others out there have had similar experiences.  As part of being a diamond level member I get access to supposedly "higher" rated analysis of 2500 level strength.  Whenever I lose a game in live chess I usually submit the game for computer analysis so that I can see where I went wrong, and what I could have done differently. There have been several instances lately, where the computer shows either one of my moves to be dubious or outright blunders, and provides a better line to show me the error of my ways.  However, when I go through the computer provided line it looks to be completely losing.  When it first occured I said to myself " no I must be missing something" there is no way a computer of 2500 level strength would miss such a basic winning line.  This has happened a few times, and today was the final straw, that led me to post this article. 

I lost a game earlier today, and submitted it to chess.com for computer analysis. Upon return of the analysis I begin to go through it, and as usual the computer pointed out several errors that I made during the course of the game. However, in one particular line it told me to move my pawn from f3 to f2.  Now I was white in this game, so I'm hoping all of you chess geniuses out there will understand the inherent problem with this suggested move.  If you cant quite get it let me spell it out for you...PAWNS CANT MOVE BACKWARDS. As you can imagine I was completely dismayed by this line provided. It has made me doubt the validity of the computer analysis altogether. As I stated previously this isnt the first time this has happened...this just happens to be one of the most blatant errors I have ever seen from the chess.com computer.

So my question to all of you is have you received flawed analysis from the chess.com computer?  And for the people who run the site, are you aware of this issue? Is anything being done to remedy it?  I will post a few of my games, that I have submitted for analysis, in which the lines provided (which the computer claims give me an edge) are completely losing or at the very least inaccurate.  If someone can see a way that I can hold the position, or validate that the computer is in fact correct, please feel free to call me a fool and illustrate the winning continuation!! Below you can find the position in which the computer recommends the best continuation for white is to move his pawn from f3 to f2.  Ill copy and paste the full recommended line next to the diagram.

38. f3f2 Rxf2 39. Rxf2 exf2+ 40. Qxf2 Qc1+ 41. Be1 Re8 42. Kf1 Re3 43. Rxe3 dxe3 )

Here is another game of mine that I had analyzed in which the computer claims I have a slight advantage, and then the line it provides shows me mating my opponent. Its clearly not forced mate, and there are a lot better continuations for black. I just cant believe after I read the analysis that the computer shows me mating, and then claims "white has a slight advantage". If checkmate is only a slight advantage then I would like a slight advantage more often!!

Now in the game my opponent played ...Qg4, and the computer claims this move is a blunder (which it is because now after Kh1 the bishop is trapped).  Here is the line the computer provides in lieu of what was played:    18... Qg6 19. c4 Rad8 20. Rd2 Bc8 21. Rc1 Bg4 22. Rcc2 Bf5 23. Rc3 dxc4 24. Qxd8 Rxd8 25. Rxd8#;  and after Rxd8# the computer says that white has a slight advantage.

There are several other games that I could post, but I think by now anyone reading this gets the point im trying to make.  Really the purpose of this article is not to bash on chess.com, but to provide constructive feedback so that we can continually improve the site, in order to help all of us improve on our chess playing abilities. I really enjoy the convenience of clicking on "computer analysis" after I finish my game, and getting some feedback on where I went wrong, but if the analysis is flawed its really hurting my play rather then helping it.

I'd LOVE to answer. Because illegal moves...well, that problem has not been an issue at all for years with Chess engines, and I don't believe chess.com wrote one of their own. Plenty of good ones to use.

My only problem is, you're showing an off game position (maybe you reached it in analysis and mistakenly shown it to be the main line? or switched lines?

Here's my issue:

Here's a link to the game:

http://www.chess.com/livechess/game.html?id=22686860

Obviously same game. Too close a position to the one you put:) But the position you placed in the diagram never happened in the game.

For one thing, in the game, by the time black's queen played itself to g5, the f and g pawns were already off the board. So you're just showing a position that never occured.

More to the point, the position you show in the first diagram, not only never happened in the game, but COULD have happened, had he played 37...Qg5 instead of 37...gxf3, which he did in the game, which should have given white a clear win (with the simple 39.Bxc5 with a won position, instead of 39.Qxf3? which reaches a lost position).

37...Qg5 is the move that, HAD it been played, would have kept the game going and would have reached the position you gave in your diagram.

So it stands to reason that you would get a line showing how your opponent could have done better (to show your mistake).

But there's obviously something strange going on with position synching... Some analysis seems to talk about different positions than the ones you think, so , if you don't mind, I'd love to see the full computer analysis you were given to the game, not just this line you put, ok? otherwise...with a position not of the game, perhaps the analysis was of another subline. If I could take a look at it, I may get a better idea and be able to help.

As for the second game, again, for better understanding what we're talking about, I assume it's this game:

http://www.chess.com/livechess/game.html?id=22643892.

It seems to me like for some reason (perhaps bad fonts?), you're confusing sublines and variations with the main lines the computer gives.  It stands to reason it would show why black can't play other lines (and show the mate in them) in sublines of the lines it gives. But then, it also would make sense it would be easily understandable because the sublines would be marked with "()" or something like that. strange.

The reason I think that is because I agree with the computer analysis in that:

18...Qg4 loses extremely quickly while 18...Qg6...well, should lose I think, but you have to play it rather well (no, there's no forced mate...)

So here too, I would like to see the entire analysis you got with the computer analysis please. I'd like to help, but it's difficult to understand what's going on with partial info.

Something else is going on, I find it hard to believe the computer analysis would give 19.c4 in the alternative line... white has a lot more crushing lines...Hmmm...  unless the full analysis you'll show (if you want to) reveals anything about what's going on here, I suggest you let someone who's running this site (such as eric...for instance) about this.

Also, dxc4 in the line the computer gave in the second game is nonsensical (well, most of the line is, but that one is just a bluner). Weird.

I've never encountered any problems with computer analysis here.

So it's two strange issues. First I wonder about the illegal move. sounds weird. second, that line seems almost completely botched. Entirely strange

Grolich can you please share with me how I can send you the analysis?  Do I have to manually put in the lines that the computer has given or is there a way for me to simply copy and paste the full game text as well as the analysis?  I would like to share with you other games of mine that I find flawed as well...I didnt want to swarm the forum with a bunch of games, but perhaps I could send them to you via a private message?

Well, I tend to delete the computer analysis messages, but, I'll find out. Just requested analysis of one of my games, I'll see how it can be sent/handled

Also with regards to the first game you are correct in saying that black never played his queen to g5...after I played gxf3 he also played gxf3, however; having said that the point I was trying to make is that the computer is saying my opponent should not have played gxf3, but rather should have played Qg5, to which my response should have been to move my pawn from f3 to f2.  Now we can debate which moves are best for white and black all day long, but the biggest thing that I wanted to bring to the forefront was that the computers suggestions of continuations are clearly flawed.  This flawed line was easy to spot because obviously you cant move your pawn backwards, but my big concern is what about some lines that are more subtle...do I need to closely scrutinize these lines for mistakes or can I take them at face value and trust in the "2500" level strength of the computer to improve my play?

Grolich you gave me an excellent idea.  Im terrible at deleting my inbox on here so I have saved every single game that the computer has analyzed for me in my inbox. Here are both games for your analytical enjoyment. Now you can see first hand exactly what im talking about.

http://www.chess.com/home/computer_analysis.html?id=22686860

http://www.chess.com/home/computer_analysis.html?id=22643892

Heh. I'm afraid these links only apply to your own user when you're logged in. Others can't access it.

Sorry. That actually makes sense. Though there should be a suggestion made a add a "copy computer analysis" button in that window, to facilitate easier discussion about specific analyses in the forums

However, pasting everything that was written in the analysis window in these messages (EVERYTHING, with the parentheses...in one go. don't paste parts and then the next part... too many potential problems) may be sufficient in this case:)

Here is the complete computer analysis from game 1...parantheses and everything :) I think this is far from an ideal way to review the analysis, but I guess it will have to do for now.

i dont know if the computer analysis of my games is flawed to say, but it suggests alternatives to my moves that put me in positions i wouldnt want to play calming such and such advantage...but the move i played led to a good position too so whatever...it never gives me any praise even though some of my sacs lead to forced mate of win of material, but its always ready to point out my "inaccuracies" and "mistakes"

unfortinatly i don't trust the computer analysis some of the games i played it give alternitive lines that are 4 or 5 pawns worse off still winning how can it say my play is inacurate with these numbers i havn't used this for a while though i might be better now

Jpd what level of analysis do you have access to?  When I was just a gold star member my computer analysis strength was 2200...now its supposedly 2500 since ive upgraded, but I feel that the analysis I receive now is worse then is was before!

i was a gold menber 2200 and 2 games a week was my limit it's like it dosn't take into accont each individual position to asess wether it is good or bad

somehow the html of this page broke. oh well.

answer to your question: all software has bugs. even the computer analysis engine we use. and unfortunately, we can't fix that one. :( i hope that you will use computer analysis for what it is: a place to generate ideas and help you see things you might not otherwise see. but it isn't perfect or the end-all, be-all of analysis :D

Hmmm... Now I'm confused.

Are you talking about the bug that gives illegal moves? plenty of engines (both free and commercial ones) without such a bug. Lots. Saying you can't fix it means it's not something "home grown" but rather existing software. Well, if it's an existing piece of software that is no longer supported, just switch software. 

Illegal move bugs are rather non existent in today's engines, so it's rather strange.

As to the html being broken, it may be due to wrong treatment of either escape codes or html tags... If so, it may be related to the last character in malurn's last computer analysis message...:)

About the html, more probably it's due to all the closed and opened html and

Erik, Since the weirdest activities provide loads of fun (especially if they take a few seconds), I made a quick check of this page's html. For some reason, There's a missing at the end of malurn's message (after the "quote" part). When you place one in artificially, the page is back to normal (you'll see two consecutive s there when there should be 3.

There are 3 divs to close:  

Imho, much stronger engines are widely available, just do your own analysis using one of them.