This has been known for a long time, but I had no idea it was this terrible. LOL
I played a game with 96.5% accuracy (including one "brilliant") and I got a rating of 1500. Hikaru plays a game with ~80% accuracy and it gives him a rating of around 2700 or above. (Granted, the game I played was only 15 moves, but still.)
I thought that was bad, but this is a whole new level.
No one takes these enhancements seriously except the marketing department.
A few years ago, game review focused on data. There were overall accuracy percentages, as well as accuracy broken down by each piece. Average move times were included. Quite a bit more, too.
The problem with data is that it requires knowledge and intelligence to use. In order to broaden the appeal of game review to a larger pool of chess players, it made sense to offer meaningless information that seems attractive. You now have a personal coach with a limited understanding of perhaps a dozen chess concepts.
Yeah also love the "AI coach" spouting nonsense. It attempts to explain a meaning of a move by detecting a pattern, but usually misses the bigger picture of why that move was important. But seriously it's not impossible to have the estimated rating to give something that reflects the truth atleast.
Yeah also love the "AI coach" spouting nonsense. It attempts to explain a meaning of a move by detecting a pattern, but usually misses the bigger picture of why that move was important. But seriously it's not impossible to have the estimated rating to give something that reflects the truth atleast.
Yes, but if you have watched streamers play "Guess the Elo", you know it's not easy for them to be accurate in many situations.
Someone at chess.com thought it might be useful for people to get an assessment of how they played in a game. Accuracy scores are hard to interpret. It seems reasonable to me that people play better or worse than their ratings in most games and that you can look at the multiple phases of the game to make a useful assessment of that play.
That's all it is. A tool to help players focus on the parts of their games that need work.
It is not deceitful. It is not a lie.
They really need to stream line the site take out all the fancy stuff and just get back to what's most important which is the chess, the lag is terrible
Yeah also love the "AI coach" spouting nonsense. It attempts to explain a meaning of a move by detecting a pattern, but usually misses the bigger picture of why that move was important. But seriously it's not impossible to have the estimated rating to give something that reflects the truth atleast.
Yes, but if you have watched streamers play "Guess the Elo", you know it's not easy for them to be accurate in many situations.
Someone at chess.com thought it might be useful for people to get an assessment of how they played in a game. Accuracy scores are hard to interpret. It seems reasonable to me that people play better or worse than their ratings in most games and that you can look at the multiple phases of the game to make a useful assessment of that play.
That's all it is. A tool to help players focus on the parts of their games that need work.
It is not deceitful. It is not a lie.
I haven no idea what you try to say with this post. Of course it's hard to make an estimation of a rating based on moves made in one game, but that's what the metric pretends to be in the chess.com game review.
But it is none of that. It doesn't even attempt to give you an accurate estimation based on your play. That is proved by the image in my opening post. It is a random number that has nothing (well, very little) to do with how you played, but is mostly just based on your actual rating. So yes, it is presented as something it is not, it is a lie and deceit by chess.com to make their product look better than it is.
This claim is completely false. Let me explain. I regularly analyze my games in liches on chess.com. In my best games I can't exceed 2500. So my game gives the same rating according to my depth of play without knowing my rating. If you analyze 3 or 5 games of top players in liches on chess.com, you can see ratings above 3000 in low accuracy. This leads to the following conclusion. The algorithm calculates these scores according to the game difficulty or depth.
This claim is completely false. Let me explain. I regularly analyze my games in liches on chess.com. In my best games I can't exceed 2500. So my game gives the same rating according to my depth of play without knowing my rating. If you analyze 3 or 5 games of top players in liches on chess.com, you can see ratings above 3000 in low accuracy. This leads to the following conclusion. The algorithm calculates these scores according to the game difficulty or depth.
The claim is true, proof is in the opening post and I don't know what youre talking about.
Try analyzing the bad games of top players from Lichess and you will see 3000 ratings in many games. The picture you posted above is mmeaninglesseaningless. It's unclear what kind of game it was made in. Maybe it's a specially prepared test. Or it could be a bug. I offer you a method you can test. If you take the game from Lichess with pgn and upload it to chess.com, you will realize that the game is not rated according to the owner.
You're missing the point that high accuracy in a poor quality game doesn't lead to high scores. Or if your opponent gifts you the queen, you will not get high points even if you play with 100 percent.
Try analyzing the bad games of top players from Lichess and you will see 3000 ratings in many games. The picture you posted above is mmeaninglesseaningless. It's unclear what kind of game it was made in. Maybe it's a specially prepared test. Or it could be a bug. I offer you a method you can test. If you take the game from Lichess with pgn and upload it to chess.com, you will realize that the game is not rated according to the owner.
You're missing the point that high accuracy in a poor quality game doesn't lead to high scores. Or if your opponent gifts you the queen, you will not get high points even if you play with 100 percent.
If you take a game, and put high ratings on one or both players, the estimated rating will be high.
If you put low ratings on the players, on the exact same game, the estimated ratings will be lower,.
The system uses the ratings of the players, in conjunction with the associated accuracies to generate the estimated ratings
Bro, you don't understand what I'm saying. If you pick a game from Lichess and upload it to chess.com, chess.com doesn't know who the game is from. I tested it that way and good players still get high scores.
Yeah also love the "AI coach" spouting nonsense. It attempts to explain a meaning of a move by detecting a pattern, but usually misses the bigger picture of why that move was important. But seriously it's not impossible to have the estimated rating to give something that reflects the truth atleast.
Yes, but if you have watched streamers play "Guess the Elo", you know it's not easy for them to be accurate in many situations.
Someone at chess.com thought it might be useful for people to get an assessment of how they played in a game. Accuracy scores are hard to interpret. It seems reasonable to me that people play better or worse than their ratings in most games and that you can look at the multiple phases of the game to make a useful assessment of that play.
That's all it is. A tool to help players focus on the parts of their games that need work.
It is not deceitful. It is not a lie.
Kind of the problem is the estimate is sometimes used based on someones rating and not just the game whenever people do guess the elo it's usualy just based on the game at least from what I've seen
It's decetful in the sense that elo is already taken into account it's just a estimate of how well you play based off of previous elo +game not just the game
Try analyzing the bad games of top players from Lichess and you will see 3000 ratings in many games. The picture you posted above is mmeaninglesseaningless. It's unclear what kind of game it was made in. Maybe it's a specially prepared test. Or it could be a bug. I offer you a method you can test. If you take the game from Lichess with pgn and upload it to chess.com, you will realize that the game is not rated according to the owner.
You're missing the point that high accuracy in a poor quality game doesn't lead to high scores. Or if your opponent gifts you the queen, you will not get high points even if you play with 100 percent.
If you take a game, and put high ratings on one or both players, the estimated rating will be high.
If you put low ratings on the players, on the exact same game, the estimated ratings will be lower,.
The system uses the ratings of the players, in conjunction with the associated accuracies to generate the estimated ratings
Chess.com probably uses this because no matter how bad one game you play is, let's say you are >1500, then your estimated rating for that game probably won't be 500 because the machine knows you aren't that bad.
Try analyzing the bad games of top players from Lichess and you will see 3000 ratings in many games. The picture you posted above is mmeaninglesseaningless. It's unclear what kind of game it was made in. Maybe it's a specially prepared test. Or it could be a bug. I offer you a method you can test. If you take the game from Lichess with pgn and upload it to chess.com, you will realize that the game is not rated according to the owner.
You're missing the point that high accuracy in a poor quality game doesn't lead to high scores. Or if your opponent gifts you the queen, you will not get high points even if you play with 100 percent.
If you take a game, and put high ratings on one or both players, the estimated rating will be high.
If you put low ratings on the players, on the exact same game, the estimated ratings will be lower,.
The system uses the ratings of the players, in conjunction with the associated accuracies to generate the estimated ratings
Chess.com probably uses this because no matter how bad one game you play is, let's say you are >1500, then your estimated rating for that game probably won't be 500 because the machine knows you aren't that bad.
If you play like a 500 you should be estimated to be 500 and a good chunk of the time the opposite happens
Try analyzing the bad games of top players from Lichess and you will see 3000 ratings in many games. The picture you posted above is mmeaninglesseaningless. It's unclear what kind of game it was made in. Maybe it's a specially prepared test. Or it could be a bug. I offer you a method you can test. If you take the game from Lichess with pgn and upload it to chess.com, you will realize that the game is not rated according to the owner.
You're missing the point that high accuracy in a poor quality game doesn't lead to high scores. Or if your opponent gifts you the queen, you will not get high points even if you play with 100 percent.
If you take a game, and put high ratings on one or both players, the estimated rating will be high.
If you put low ratings on the players, on the exact same game, the estimated ratings will be lower,.
The system uses the ratings of the players, in conjunction with the associated accuracies to generate the estimated ratings
Chess.com probably uses this because no matter how bad one game you play is, let's say you are >1500, then your estimated rating for that game probably won't be 500 because the machine knows you aren't that bad.
If you play like a 500 you should be estimated to be 500 and a good chunk of the time the opposite happens
Sometimes, people just want that little bit of validation that they played good, estimated elo is a good marketing tool on chess.com's part
"Lie" and "Deceit" are some pretty strong words...
It is what it is lmao. You can't present a "game rating" that "gives an estimation on the players rating based on the game" and then actually give a number that is mostly based on the players actual rating.
It's an estimate. It's really hard to judge a player's rating on the basis of one game, just using accuracies based on engine matching (the shown accuracies aren't even direct engine matching).
So many variables go into hane quality that basing the process on existing ratings makes sense.
I don't really pay attention to Game Review rating estimates, for this reason. It's simply not correct. I wouldn't call it a "lie" though - I'd say that it simply doesn't work the way it's intended to. It'll probably improve in time.
Also, Game Review sometimes recommends inaccurate moves. It also sometimes marks top moves as "inaccurate", too.
I suspect this is because it operates on a faster, lower-depth engine setting.
Either way, I recommend analyzing with the engine on higher depth in Analysis Mode, if you want stronger help with your chess (for the more serious players).
To me, Game review is more about being user-friendly and encouraging.
Chess.com game review presents a "game rating" for the particular game you're reviewing. The description is: "This gives an estimation of a players rating based on a single game".
That is a lie. The estimation is not based on the game, but it's actually most based on the players actual rating playing the game. The same game by a 1500 or a 2500 yield a different "game rating". It is therefore just deceit and the game review either can't or wont produce an accurate estimation only based on the game. It's laughable and deceitful to present it as such. Below an example just how ridiculous it is.
A 1500 could play a full game at 99% and they wont get such a high estimate.