Game Review Game Rating

Those ratings are worthless

Worthless is maybe an overstatement. One just has to keep in mind that isolated games means nothing much, also that shorter the game, lesser the importance of the said rating evaluation.

Perhaps there is a thing to do: evaluate your 20 last games and find the average (x/20), then see how far it actually is from your current rating. Maybe this would work better with a higher number of games, idk.

For one thing I'm lazy to do it, and then I can't really do that unless I do restrict myself to one game a day, cos free accts can't get more than one game a day evaluated...

One other thing as well, is that the formula used to estimate the Elo of the two players requires at least one of the players to have a rating. The PGN for the games played on Chess.com will contain this information in the header (the tags [WhiteElo "1000"], for example.
When it estimates the Elo for the game, it factors in that information. You can easily test this by just taking one of your games, download the PGN, and put higher or lower Elo values for the two players. The estimated Elo will go up or down, accordingly, showing how the estimate is not based solely on the moves played, but also includes information about the current Elo of the players.
As a guide to how you and your opponent played, I think if the estimated Elo's are higher than your actual Elo, you could take that as you played better than expected, and if lower, then you played worse than expected.
Given I can only do one a day, I've not played around too much with it, but I could imagine a game between two 800 rated players an estimate higher than 800 (say, the winner scored at 1000, so played better by 200 rating points) could result in a lower (but still higher) estimated Elo if one changed the players' Elo's to both be something like 2500; I could see that ending up with something like estimated Elo for the winner at say 2200 - so higher than the initial 1000 given to the 800 (you played better than expected) but indicating that the same game by a 2500 was played worse than expected. However, I don't know if that's true - it's just a possibility I've wondered about.

[Event "Vs. Computer"]
[Site "Chess.com"]
[Date "2024-05-21"]
[White "mdsheikhforid"]
[Black "Martin"]
[Result "1-0"]
[WhiteElo "800"]
[BlackElo "250"]
[TimeControl "-"]
[Termination "mdsheikhforid won by checkmate"]
1. e4 e5 2. f4 d6 3. Nf3 f6 4. Nc3 c5 5. Bc4 exf4 6. d3 Bd7 7. Bxf4 Bf5 8. exf5
a5 9. O-O Qc7 10. Re1+ Kd8 11. Qe2 Qc8 12. Qe8+ Kc7 13. Nb5+ Kb6 14. Qxc8 Be7
15. Bxd6 Bxd6 16. Re6 h5 17. Rxd6+ Nc6 18. Qc7+ Ka6 19. Na3+ Ka7 20. Nb5+ Ka6
21. Nbd4+ Ka7 22. Nxc6# 1-0

Hello

[Event "lakshay12344555n vs. mubinaa21"]
[Site "Chess.com"]
[Date "2025-06-29"]
[White "lakshay12344555n"]
[Black "mubinaa21"]
[Result "0-1"]
[WhiteElo "1014"]
[BlackElo "1008"]
[TimeControl "600"]
[Termination "mubinaa21 won by checkmate"]
1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 Bc5 5. Nxf7 Bxf2+ 6. Kxf2 Nxe4+ 7. Ke1
Qh4+ 8. g3 Nxg3 9. Rg1 Ne4+ 10. Kf1 Qf2# 0-1

As I wrote many times, ratings given by game review are in most cases completely nonsensical. Just ignore it.

Hi

The rating itself has little value. It essentially is the rating you would achieve if you were to play chess with that level of accuracy EVERY game.

The idea is that the game AS PLAYED would be statistically similar to what is expected by a player of that rating.

You rating can easily become inflated if you end up in a line where best move becomes obvious thus raising your accuracy score, and the analysis engine giving you a higher 'game score'

If the computer says that a move is wrong, that's because it has at least one refutation that makes your position worse. The computer does not make mistakes. We do.

Despite what is commonly thought, Game Review is not a better or deeper form of Analysis. It has an entirely different function which is to provide immediate feedback in a way that is easy to understand and, more importantly, that users generally enjoy. The move categories, "Great" and "
Brilliant" are entirely bogus, most especially the latter. Computers cannot tell what is "Brilliant" for a human since that would have to mean, by very definition, that it not be easy to find. Trying to replicate this, developers at first tried flagging moves that were game changers and that the computer did not recognize as "Best" before a certain depth of ansalysis. Sounds great but turned up far too many odd results and overlooked far too many moves that, by appearances, should have qualified. So in 2022, this way of calculation was replaced with the current formula: to be classified as "Brilliant", a move must now be important to the game, be the "Best" move and also must include a sacrifice. The results are better but still not very meaningful: there is no reason to think a brilliant move has to contain a sacrifice and, worse, a "Brilliant" move can be completely obvious, a contradiction in terms. But the category remains very popular and entire threads often appear with the intention of providing a place to share and compare them. The same is true of the Rating Estimate feature. Unlike a normal rating system, there is no algorithm that suffices, except in a very rough way, to measure such a thing, least of all for a single game. To say the result must be taken with a grain of salt is putting it mildly. Calling it utter BS is probably more like it. But it remains very popular, not least of all because it tends to err, predictably, on the side of generosity.

"Oh, you say I played just like a 2000?"

@SicilianSwiftie advises, forget Game Review, forget Lichess, just install "en croissant".

https://encroissant.org/download

for Linux, Windows or MacOS