What They Don't Tell You About Accuracy

You can do it yourself here:
https://www.chess.com/analysis

But it looks like you're right. I tried ratings of 500, 800, 2500, and 3000.

The percentages (mostly) don't change, but the number of inaccuracies, good moves, etc do change.

-

-

Ah, thanks. That is quite interesting. So it is only the labels.

llama, i had a short conversation with ziryab about this (i had a similar topic). i think the lesson (no pun) is to not trust any of cc's analysis. 

i had played a 5 or 10 minute game, and the analysis afterwards showed my play almost perfect. i'm ~1200 in those types of games. no chance in heck i was so "good".

Mind showing the game? 

I think analysis is far from perfect but there is a lot of good information sometimes to be gained.

My head hurts trying to understand this

@llama47 -- thanks for (a) your OP, that's amazing.  And (b) For this post, too -- that classifications are dependent upon rating.

(I see that the classifcations for 500 and 800 are the same as each other, and so too for 2500 and 3000.  So, we know of two classifications.  Have you (or can I talk you into) trying also for, say, 1200, 1600, 2000?

not entirely

I have a really bad habit of torturing myself over every small mistake I make, so it really helps me see how bad im doing and how i am supposed to be better

Usually, the problem (for U1500 folks) is missed tactics.  For that, the engine is great.  Sometimes, the engine catches bad moves that lead up to tactics way down the road (e.g., unnecessarily moving one of your pawns in front of your castled king).  But other times (as @llama47 just showed) it tells us "good moves" one moves that aren't so good.  (This is even more true in the openings).

And so: using the engine to catch bad moves even distantly related to tactics -- great.

Using CAPS -- usually decent but add a lot of grains of salt.

PS: easier said than done, but: don't torture yourself (well, unless you like too ).  We *all* make mistakes.  Look at each game as an opportunity to learn from those mistakes.

Someone who thinks accuracy is mostly worthless is deluding himself. It's far from perfect but it gives a decent indication in some instances where the opening and forced moves don't play a big role and to compare your own average accuracy over time etc. 

lol thx

Its basically my method of coping with my ego

whenever i did a test in school, i never used to double-check it, and then i got it back with a low C

so i just make each mistake seem like the end of the world to myself so that I don't make any

gotta say its working really great in school, not so much in chess

21 good moves supposedly actually. and 1 book move.

I've been wondering about game accuracy vs game quality for a long time, and I agree that a high accuracy game can be low quality, while a high quality game may only have a good accuracy score. An example of the latter is Radjabov vs Carlsen, Baku, 28/04/2008. It's in the Sicilian Dragon opening, and has lots of tactics. The accuracy scores for White and Black are 80 and 86. I now think trying to understand the mistakes pointed out by the game review is more important than the accuracy score by itself. They are useful tools.

yea gm games have really low accuracy scores. The way accuracy scores are determined is by comparing your moves to the engine at a certain depth. Good and Excellent moves are always good in that position unless there is a much better or obvious move that wins. (i've had games where hanging your queen in a completely losing game is considered a good move, so take the accuracy with a grain of salt). The games lose accuracy based on the length of a game, because the engine prefers different moves than what a player sees as best, and the difference between excellent and best moves is simply preference. Thats why some top players play dubious openings, simply to mess with their opponent because they know that their opponent didn't study that opening

wow