@Rogue_Berserker said in #25:
> You have misunderstood this. To effectively identify the vast majority of cheating,
Chess.com computes an aggregate Strength Score. Strength Score is a measurement of the similarity between the moves made by the player, and the moves suggested as “strongest moves” by the chess engine. In a sense, it is a measure of the accuracy of play. In short, it's pretty much the same as accuracy over several games.
I see you have read (and not quoted) "The Hans Niemann Report" by
chess.com. Have you also read the bottom notes?
Particularly (18):
> Though notably, Strength Score is calculated differently from the “Accuracy Score” shared with a
Chess.com player when they review their games. In essence, Strength Score is based on actual statistical models and meant to be used across multiple games, while Accuracy is a product-driven score meant for one game, using a different, and less statistically-driven algorithm.
In #23 you are not referring to the Strength Score, because how could you since it's not public data, but to the game review accuracy score. Which is misleading, not just because it is calculated differently than the Strength Score (they say), but because we don't really know what it actually is, since "the math" behind it is not public:
> Previously, CAPS (v1) looked to create a 0-100 band within the normal human player range. So, scoring perhaps 40% “Best” moves, which is very low, was equal to single-digits on CAPS. And, scoring a very high number of “Best” moves, but not a perfect game, was often rated 99.9 on CAPS, even though it wasn’t played perfectly.
> This made some people feel bad (on the low end), and led to a lot of cheating allegations (on the high end).
> You will notice that the majority of scores now fall mostly be between 50 and 95, which provides a more intuitive understanding of how accurately you played in your game.
Basically, in poor words it says that, unless you screw up very very badly in the game, your accuracy will always be starting from an average of 50. That is to meet lower rated players' dissatisfaction and complaints.
In a sense we can say that based on your rating on the ladder you will receive a more or less accurate accuracy.
On this matter Lichess states:
> While there is some correlation between the players' ratings and their accuracy, it is not straightforward.
Because:
> A more skilled player tends to play more principled, theory-heavy openings and put more tactical pressure on the opponent. This can create more complicated positions and provoke inaccurate play on both sides. Moreover, lower-rated players are often more reluctant to resign. As we discussed above, protracted lopsided endgames can increase the accuracy score.
Moreover, while chesscom thinks that:
> The new Accuracy scores, based on CAPS2, replicate the feeling of being graded on a test in school.
Lichess states:
> It is flawed to compare accuracy to a numerical grade you would get on a test.
Because:
> In more complex positions it is harder to find the best moves, so your accuracy might drop accordingly. Conversely, in lopsided positions most moves don't change the winning chances meaningfully, so the accuracy score may be high even if your conversion of the position wasn't clinical.
The HN report states that:
> Rating plays no part in
Chess.com’s Strength Score, as players can significantly over-perform or
underperform their rating.
Now, to answer your question in #1:
> A very high accuracy percentage isn't necessarily indicative of superhuman, "GM-level" play. If you blunder early on or play consistently subpar moves, your opponent will have greater chances to capitalize on your mistakes. This can lead to a very high accuracy.
#2 is a TL:DR of all this.
Sources:
www.chess.com/blog/CHESScom/hans-niemann-reportsupport.chess.com/article/1135-what-is-accuracy-in-analysis-how-is-it-measuredlichess.org/page/accuracy