Is accuracy, a joke?

As long as you don't make a bad one, everything is ok.

26/28 is 92.86%...but i'm pretty sure they have a formula for this, meaning that perfect moves and a good move are not calculated and worth the same. There wasn't any blunders or missed wins so 98.1% accuracy isnt too bad

Of course a second best move is less accurate than a first best move, and a move that loses 1 point of evaluation is more accurate than a move that loses 2 points of evaluation

Definitely one of my stronger games, but an outlier.  While I do occasionally see best move %s in the 60-70 range, I estimate my average best move% to be in the 48-52% range (that's based on a small sample of 30 games over the last 3 years).  And on the flip side, there are also occasional games in the 30s too.

I really want to better understand what accuracy is, how to read it, how to use it to improve, etc.  I currently lean on best move% because 1) I understand it, and 2) GM Hess talking about (paraphrasing here) focusing on finding the best move in every position vs. playing for results really made sense to me.  I keep thinking that if I understood accuracy, it might be an even better tracking metric than best move%.

- M

The computer shows Be3 as the best move with evaluation +.37, with Bc4 evaluated at zero.  So I think good is defined as within a certain range from the best move evaluation, even though it wasn't listed as in the top 3 moves.  Bishop on c4 in the opening is standard position in the Italian game, so it isn't a bad spot for it.  

No offense taken at all.  I know exactly where I stand as a beginning player, and I agree with everything except the "worthless opinion" part...

- M

Just a thought…

The other day, I heard Sam Copeland talking about S-fish evals vs Komodo evals – (paraphrasing here) that 1) the former might eval a position as +1, where the latter will eval it at +0.5 and that 2) neither is necessarily correct/incorrect,  But the thresholds for the two differ in how the evals are presented.  BOTH are correct, but the end users have to interpret the two “different” evals with the knowledge that they’re essentially evaluating a position similarly, but presenting it differently.  So that got me wondering…

Is it possible that the accuracy #s presented on the game reports are “inflated” relative to other metrics?  In the same way that S-fish evals are “inflated” relative to Komodo evals?

In other words, in a game with a 50% best move%, and an accuracy of 98.0%, that BOTH are correct, but one looks “better” because the threshold for the accuracy values are “inflated” relative to the best move% (and/or other metrics).

In other-other words…  should chess.com lower their evaluation thresholds for the accuracy scoring so that the end result presents a “more realistic” number to end users - eg, something that more closely reflects the best move% (or some other more commonly recognized metric)?

- M

It's my old account's game!

OMG they killed Kenny.  Those... 

Still upcoming? 

I think the formula should take the actual numerical difference from best for each move, rather than general values for different categories of moves.  The computer will call a blunder for a minus 1 and then call an innacuracy for a minus 3 later in the game, so I wouldn't use the computer categories to calculate accuracy.

Accuracy is really just a reflection of % best moves and the ave difference value for the moves that weren't best.  The problem with using ave diff is that one blunder (like blundering a queen and being minus 12 evaluation from best) skews the whole result even if the rest of your moves were very accurate.  

Just out of curiosity, as this may be to my 6-700 avg level, but has anyone seen or made a "brilliant" move here?  

https://www.chess.com/forum/view/game-analysis/brilliant-moves-in-new-game-analysis-report?page=1

Accuracy IS subjective. Don't believe them engines!

Thanks Forked_again!

Who has been the most accurate player of all times? I mean...Elo system could not be appropriate method across centuries to value the real straight of one player due to different conditions ( straight of opponents, number of games i,e) but Accuracy could give us the real straight of a player. Who is the best Chess Player ever?..Check every single moves the played over their times and calculating their Accuracy % rated by a professional chess software and we will have the answer...

They did that analysis several years ago for a CAPS article.

https://www.chess.com/article/view/who-was-the-best-world-chess-champion-in-history

Hello, I find the engine to be helpful with going over the game, but not completely with the actual score itself. The computer just has a different strategy, that doesn't make our moves worse necessarily, no? However, I am posting because my brother and I compared our scores and found the engine have us different scores. So how accurate is this accuracy thing in reality? Seems like a hot air balloon ride when you're being told it is a helicopter.

Issue
The issue is the analysis tool made available after every game. 
https://www.chess.com/live/game/6379722028
A recent game where the result was reviewed. 
28 moves with an accuracy score of 97.7%
The “Report” gives 3 book, 18 best, 5 excellent, 1 good, 1 mistake.
The “analysis” gives 3 book, 14 best, 9 excellent, 1 good, 1 mistake.
A discrepancy of 4 less best moves and 4 greater excellent moves (not the best computer move).
The report differs from the actual game analysis.
The accuracy score is defined as “measuring the accuracy of your move on a scale of 0-100 when compared to top computer moves.”
The report gives a differing total of best and excellent moves then seen when reviewing the game.
Of the 28 moves - 17 were top computer moves. 
11 were inferior moves including a mistake.
How on earth is a 97.7% score achieved ???
Especially when the “report” and the “analysis” differ in the reported numbers.

also depends on the deph of stockfish , i use 35 and than the accuracy drops to something around 87-89 %