What accuracy % do you consider a “good” game?

 Because they dont have to. And the better their processing power / evaluation algorithms, the less they will need to understand those concepts.

  Those kind of concepts are just abstractions we humans create to deal with complexity beyond our understanding. 

I will also say that computers like Stockfish don’t understand that the person who is down a rook but has a mating attack like 7 moves before mate is winning. Until the computer sees that your follow-up is actually deadly, I seriously imagine the computer doing this: Dude why did you play that? Now you’re down a rook, and you will lose so hard! You are a loser to not listen to a computer. Then the human does the follow up and the computer says this: Oh I’m actually stupid. I didn’t realize that was winning.

LOL

I am a 700 and I played a game with 96.4 accuracy

id say at least 80% is something id be proud of

The way I see it, accuracy is a measure of how close you are to playing like Stockfish. I get pretty paranoid at times, and if it feels like every move my opponent makes is going to be the right move, then I'll play out the game for as long as I can and then run Game Review to see how close to engine play it was. This only makes sense, though -- if it makes sense at all -- because I'm stuck in the high 600s / low 700s. In a game with 35 moves, how is someone rated 650 getting a 95% accuracy? I mean, it happens. All the time. But it seems like a red flag. I think the analysis here is pretty thorough, but one thing I discovered via ChessBase was that this is actually two different types of analysis -- tactical and centipawn -- and the latter is apparently designed to catch people who aren't playing fair. So I'm not so sure it was ever meant to tell us how well we played.

stockfish actually also analyzes deadly attacks like this regardless of the material advantage/disadvantages as long as it has enough depth

85&

85%

Neither side hanging pieces or tactics makes it a good game to me.

Let's see. "good" is better than terrible, and worse than amazing. Unless the amazing is for amazingly bad. So now all you have to do is define terrible, bad and amazing, and there's your answer. Should be easy enough.

I'd say more like 85.2314551263 % or thereabout

For me, I usually like a game when I do other thinking than just "dumb calculation". For example a strategic idea that is not just 100% copy of something I read in a book, but there is just a little extension on it. I find though, that chess can be mostly about dumb calculation, meaning you typically get the best results with a "shut up and calculate" approach. Given a few basic strategic ideas, I find it is quite uncommon that real "outside the box" thinking gives good results.

@woloexe true, but once at chess camp we were analyzing a game, and Stockfish didn’t see the win until the follow up was played.

Sorry, woloexe was commenting on page 6 of this forum, I know this is page 7, but I couldn’t respond until now.

If you’re wondering my average accuracy in rapid games(10 min) it is 75.19 on average. It’s an average of 78.6 when I win, and like 72 when I lose.

I’ve had games that are like 90% accuracy, but I’ve also had games that are like 35% accuracy when I play bullet, because you have to move like Sonic 💨 

Anderssen and Kieseritzky shouldn't take it too much to heart.

This is a KPPvKPP enddgame played by Nalimov v Nalimov. Every move is perfectly accurate.

Coach doesn't think much of Nalimov neither.

(Incidentally - I don't think the chess.com engine has a % feature, just a CAPS2 score which is not a % of anything.)

Well depends, higher accuracy can be achieved based on the strength of your opponents moves. For example, if your hangs their queen on move 4, you can just proceed to trade everything and achieve 90%+ accuracy.