Chesscom's Game Review Guess the ELO Feature is Bullsh*t

[@scottb32a]

I only take notice of the rating when its says Im great , otherwise I just assume its wrong

[@ChessBootCamp]

I think that's a very sensible approach.

[@johntate8561]

Well done for calling this out Ben, I have long had my doubts

[@ChessBootCamp]

Cheers.

[@theinquisitor5518]

Game 1: I guessed 1300 for both players Game 2: I guess 1400 for you and 800 for your opponent I agree the ratings don't "feel" right. But I think the ratings depend a lot on how "obvious" the moves are. In game 2 you made all the obvious moves which made you think you didn't play well but all the obvious moves turned out to be the best ones which is why your rating was so high even though you thought game 2 was a bit meh. Game 1 felt more sharp and so it was much harder to find the best moves.

[@ChessBootCamp]

Yeah absolutely. I think what we’re saying is, it’s no indication of the quality of the chess game, and therefore trash data.

[@FLSharks]

It also said the 2nd guy had 0 blunders even though he let that rook go

[@ChessBootCamp]

Yeah, just hard to explain.

[@tfosss8775]

The guess the elo feature definitely takes your and your opponents rating into account when determining elo for the game. For Example, on LiChess I am 1563, had a game as black, against a 1543. I copied the PGN from LiChess into Chess to analyze the elo, I did this 3 times each time changing the ratings of the 2 players. When I plugged in our real LiChess ratings 1543 (w) v 1563 (b), Chess elo said we played like 1050 (w) and 1900 (b), almost -500 (w) + 340 (b) When I changed the ratings to 1043 (w) v 1063 (b), Chess elo said we played like 950 (w) and 1600 (b), almost -100 (w) + 540 (b) When I changed the ratings to 2043 (w) v 2063 (b), Chess elo said we played like 1600 (w) and 2150 (b), almost -450 (w) +90 (b) It did give us the same accuracy score all 3 times, 69.3 (w) and 81.3 (b)

[@ChessBootCamp]

Oh yes, that's well known. And I personally think it's dumb. Is it impossible for a 1500 to play a 2500-level game? No, but it is on chess.com.

[@cedriccappelle2036]

I think they can make a more sophisticated guess the elo, but it would need so much computation that it's not worth it, or they need to limit the uses per day, even at paid access. The "simple" solution would be to let the computation be done on the user hardware, but that also has issues. Lastly, guess the elo is sort of a meme, so I think it's fine that it's way off now and then! 😄

[@ChessBootCamp]

Yeah, it’s basically garbage, “For entertainment purposes only.”

[@francescoragghianti6068]

Yeah I noticed it's trash too. I checkmated my opponent in 4 moves on a good game I won in a few moves. It wasn't a forced mate, but since they were four moves where I could have gotten a free pawn they were four missed moves, so my guess the Elo went down badly. On another game my opponent played almost like you second game but he had no missed moves, his guess the Elo was greater than my previous game

[@ChessBootCamp]

Ya, it really is just nonsense.

[@thorsthunder2670]

Its fundamentally a problem with the way we calculate accuracy percentages, which is comparing the evaluation of your move compared to the evaluation of the computer's move. The more sharp a game is the lower the accuracy is going to be, that's simply a fact when it comes to human chess, in a boring position grind the players will always be playing closer to the computer than in a wild swinging King's gambit. Famously when world champions were compared the computer, Capablanca, (known for his positional, grinding style) ranked higher than Tal (known for his wild sacrifices and attacking), despite Tal being world champion 40 years later. The reason the second game is so much higher is a different phenomenon. Because the blunder came early, 3/4 of the game (in terms of moves) was played with you having a massive advantage. When you're -5 and converting the game it doesn't really matter if you play the top move or the 5th move, the evaluation will remain pretty much the same and the computer will mark it as highly accurate, same for your opponent, his moves can't really move the dial much. The guess the elo system has always been nonsense since it bases it off of your current elo. The same game played by a 1000 and a 2000 would come out as completely different guesses. At the end of the day they're amusing to look but not a useful evaluation of chess level.

[@ChessBootCamp]

I agree entirely, thanks for the input bro.

[@cedriccappelle2036]

Challenge accepted. My guess for game 1 is 1300. Auch, I'm far off. Second game there is nothing to evaluate. Opponent blundered, and then put 0 resistance. Perhaps 1000-1300 for you and -500 for opponent. Alright, it makes no sense 🤣

[@ChessBootCamp]

Ha ha, told you 😎

[@AtlasRanger]

900/1000 800/1500

[@ChessBootCamp]

So close! :-)

[@rojavida]

Game 2 - I have you at about 1400 no obvious mistakes, but some “interesting” decisions. Opponent only about 800. Some appalling moves by them.

[@ChessBootCamp]

Yeah, if anyone guesses within 100 accuracy on all of them my final flabber will be gasted.

[@byronwindhorst3220]

I averaged all my games and came out with 1100 ELO. I'm still at 1008. I really should start playing again, but the videos of other games are fun to watch. By the way, where do get the goofy images for the icons for the videos?

[@ChessBootCamp]

I have a premium subscription to a resource, vecteezy I think that one is.

[@TrentHrappstead1]

When reviewing, play against the computer for the last move you played and stop the game after that. Re analyse the new game against bots and the game should be rated more accurately

[@ChessBootCamp]

Wow that's interesting, might have to try that!

[@joseraulcapablanca8564]

first game im guessing 950 or so each. Lovely mate by the way Ben. Second game i am saying 1150 from you would have been a hundred or so higher if you had made the correct capture on c6 and a hundred or so higher had you not played the c6 defence. Him i am gonna give 450 because you know how generous i am whne it comes to chess. That was bullshit. I do like the game review facility. But that is utter bullshit 1750 and 2200. Thanks mate.

[@ChessBootCamp]

Yup, it really is ridiculous. I know they have to have some way of guessing, so I'm not saying they're doing a bad job, just maybe it isn't even worth doing. It's a bit like people who say, "I had a good walk today, I did 15,000 steps." But did you ENJOY the walk??

[@niravapurv4578]

My post got deleted becuse its not allowed to write the name on a sertain webpage.. i write a shorter one. I like your comparison of these to games. .and i suggested before that the review from a sertain webpage giving suggestions on what kind of elo level you played at is kind of faulty or at least supperficial.. its like they prolong the accuracy score to give a elo number... but when you see the accuracy score you understand that is based on numbers of blunders, errors, misses and inacuracies you played on a game.. or even better the average centi pawn loss per move.. The thing is the comp dont see the difference on what is a stupidly hard computer line that only the best chessplayers can see.. and a move that blunders a queen in one move.. and this is clearly illustrated by your comparison here.. the last game has less blunders, but the blunders are more obvious and easy to spot for a human.. and the computer dont understand that and give a high score bcause the game was less tactical/sharp, and so there was less number of blunders.. where the first game there was lots of errors but they are not easy to see unless you are realy good.. but computer see just lots of blunders and misses... So stop looking at that feature so much. They might make it better with ai in the future.. they should.. but the webpage you working on is not givving value for the money, so it might take a long time before they make that better, the ai need to be improved for that. and its going to take time. Also i want to say I stopped playing on that page some years ago, i went to lichess and there is no way back. I am scared this webpage is gone buy lichess. but i hope and think that is not going to happen. because if so they gone destroy it and i have probably more than 50 studies full of good chapters. I have only like 500 games on that webpage, but i cant even see most of my games because they are some years ago, and thats a no-no for me. i want to see and use the first game i saw and got traxler counter strike thrown at me for instructive purposes. Because i lost that game and its a good instruction on how and how not to play in that opening. But that game i have to pay to see.. my own games, i have to pay to see. that is rubbish to me. I recomend to move to lichess, the game review is good and after game and game review, before you go through the whole game you can use the "find a better move"/"learn from your mistakes" feature.. its very nice and as far as i can see the other webpage dont have that. I never paid there, so i dont know the full potential on that webpage, but i went to lichess and never regret... no way back for me.. perhaps its better if paying, i might descover more of that later as i would like to acces some of my arealier games.. but how do i secure acces to those games for the future.. i have to copy all of then into studies at lichess? i dont know.

[@labestianegra6373]

You want a comment? Okay. You look like Jamie Carragher.

[@ChessBootCamp]

Had to look him up but OK, I'll take that.