Forums

Bug with "Game Rating" in Game Review

Sort:
BigNose

My opponent played a terrible opening, quickly losing 2 pawns, and then resigned at move 16. But the game review gave him a game rating of 2400!

Here is a link to the game review:
https://www.chess.com/analysis/game/live/123103876715?tab=review

This seems like a bug with Game Review and it would be great if it could be fixed to be more accurate.

Martin_Stahl
BigNose wrote:

My opponent played a terrible opening, quickly losing 2 pawns, and then resigned at move 16. But the game review gave him a game rating of 2400!

Here is a link to the game review:
https://www.chess.com/analysis/game/live/123103876715?tab=review

This seems like a bug with Game Review and it would be great if it could be fixed to be more accurate.

At maximum it's 2350 but the algorithm uses both player's ratings and associated accuracies to generate the values. Since it gives you a 2500, the opponent estimate seems to be in line with the algorithm.

BigNose

It should not take into account the players rating, only the moves played. His opening play seems at most 1600 to me, so it does seem like a bug.

BigNose
Martin_Stahl wrote:

At maximum it's 2350 but the algorithm uses both player's ratings and associated accuracies to generate the values. Since it gives you a 2500, the opponent estimate seems to be in line with the algorithm.

Is there a document or web page that outlines the algorithm?

Martin_Stahl
BigNose wrote:

It should not take into account the players rating, only the moves played. His opening play seems at most 1600 to me, so it does seem like a bug.

It's really hard, maybe impossible, to take a game and give an accurate rating estimate based just on move accuracies that will work with all games qualities across all ratings and rating differences between players.

Opening knowledge does not necessarily equate to rating either

Martin_Stahl
BigNose wrote:
Martin_Stahl wrote:

At maximum it's 2350 but the algorithm uses both player's ratings and associated accuracies to generate the values. Since it gives you a 2500, the opponent estimate seems to be in line with the algorithm.

Is there a document or web page that outlines the algorithm?

No, that is not public information.