Look at Yifan's accuracy, how can this be?

I could understand a top 10 in the world player, but isn't Yifan lower than the 2700s in the world? If she is getting 98% and 99%, are the 2700+ players playing with 105% accuracy? How does this work?

Chess is a game that requires two people to play. Higher rating denotes lower probability of making mistakes. Accuracy depends on the position and opponent. It is easier to take advantage of a noob who blunders all the time, instead of a master who gets confused by a situation. Conversely, it is less likely for major mistakes to occur in quiet positional games like the Giuco pianissimo, instead of tactical shootouts like you would see in the Botvinnik variation of the slav defense or many variations of the KID.

That doesn't answer my question. I am not asking why is a high rated player's accuracy high, I am asking HOW ARE EVEN HIGHER RATED PLAYERS EVALUATED AS IT PERTAINS TO ACCURACY?

All you did was say higher rated players play better. Yes, I know that. Richer people buy more expensive houses and cars. We can all surmise that.

What I am asking is if Hou Yifan scores 99%, then how does a 2750 score? 105%?

If she is giving an equal fight then shouldn't she be rated higher? Or is this the reason? Her chess.com rating is lower than her true ability, so when the server evaluates her play it is actually handicapping her because she hasn't played as many games on here?

No this is percentage so it cant get above 100% in this case.

Hou yifan is a super gm so it makes sense. Also ratings don't really matter too much because people are playing not their ratings

So most super gms play as good as yifan

yep or else they wouldn't be super gms

Since it is not possible, I am wondering if she is being handicapped. What if Carlsen played the same game, would he have had 99%?

The accuracy score is a one-dimensional number that only tells a small part of the story. It's not too hard scoring 98%+ accuracy against players who are out-rated by a large margin.

In that list of games (above) Yifan only scored 97% or more ONCE against a player rated 2500 or higher... all the other 97%-to-99% scores were against players rated 2200 to 2400 or so. It's not to difficult playing accurately if your opponent isn't capable of putting you under serious pressure.

To answer your question about Magnus and other top-10 Super-GMs, they could rack up impressive accuracy scores against other 2700+ Super-GMs. That's a much more difficult feat.

"To answer your question about Magnus and other top-10 Super-GMs, they could rack up impressive accuracy scores against other 2700+ Super-GMs. That's a more difficult feat."

No, what I am asking is the SAME opponent. 

Let's say Person A is 2600 and they play PERSON B 2600, it is a fair fight. Equal.

Now, let's say Magnus Carlsen plays PERSON B. How would he perform if he is playing at his 2800+ rating? Probably better, right?

Ok, now let's say Person A signs up to chess.com and has a rating of 2400 but if they are 2600 OTB, then it should be higher, right?

Ok, now, we can put the info together to get the question I am asking. 

Is the 99% because the 2600 player is rated 2400 on chess.com? YES or NO. There is no "Well this well that, if you look at the 2200 player blah blah malarky".

If the answer is no, then I would ask, what percentage would Carlsen get, 105%? Since that appears to be impossible, my only guess is that Yifan's 97%+ scores are a result of her lower chess.com rating.

2/10.  Your trolling career is not looking good.

Since no one has even attempted a serious answer here, no of course above 100% is not possible.  At least two things are possible.  First, the percentages aren't accurate to the level of precision quoted.  Second, at the very highest level, a person who is 98.8% will eventually lose to a person who is 98.9%.  Or some other factors.  Or some combination of these or other factors.

No.

If the accuracy score here on chess.com has any relation to a CAPS score (which it seems to be modeled on), then it is derived from the cumulative difference in centipawn value between the moves actually played and the computer's theoretical "best" moves.

To put it another way, the fewer "dropped" centipawns over the course of the game, the higher the accuracy score.

A super-GM would score only fractionally higher against the same over-matched opponent.

This should be obvious if you consider a thought experiment:

Suppose Magnus plays against some dummy who blunders his pieces one after another and then walks into a mate. Magnus would get a very high caps score, because it isn't hard to find the absolute best moves when your opponent makes one blunder after another.

If I were matched against the same blunder-prone dummy, I could get almost the same accuracy score as the World Champion. I can spot a free Rook or a mate-in-2 almost as easily as Magnus.

It's the accuracy score against opponents who are ALSO playing accurately that distinguishes between super-GM play and my own thud-and-blunder style.

Ok, so, let's say we got Carlsen completely drunk (to the point where he was vomiting and saying weird things like "I'm gonna whoop that tiger of Madras and become world champion" (Thinking it was 2013 again), could an 1800 player get lucky and beat him? If so, wouldn't their accuracy be higher? Or is that not how it works?

"If the accuracy score here on chess.com has any relation to a CAPS score "

I have no idea what you mean by caps. Can you start from scratch and explain what makes an accuracy 99% when there are super GMs hundreds of points higher? It doesn't make sense to me.

Not caps. CAPS (in capital letters). Computer Aggregate Precision Score.

Computers are not gods. All they do is crunch numbers. Bear that in mind.

The way a CAPS score is determined is this:

The computer evaluates the position, decides which move it considers "best", and assigns a numerical score to that move... let's say, +1.17

Then the computer evaluates the move that was actually made in the game, and assigns it a numerical score... let's say, +1.09

Note that the move actually made cannot score higher than the computer's theoretical "best" move... or else it would BE the best move. At best, it can be equal.

Then the two values are subtracted, and the discrepancy (in this case, -0.08) shows how far "off the pace" the player has fallen by playing their own move instead of the computer's theoretical "best" move.

Go through the entire game adding up the discrepancies, and the total cumulative centipawn loss is then scaled from 0-to-100 and that is the accuracy score. NOTE that the accuracy score is NOT simply 100 minus the cumulative centipawn loss, since that could give you accuracy scores less than zero. The result gets scaled by an algorithm.

https://www.chess.com/article/view/better-than-ratings-chess-com-s-new-caps-system

Be aware that the accuracy scores on chess.com are not worth basing any speculations on... the engine search depth is much too shallow to find the "theoretically best move", and (probably for the same reason) the evaluation algorithm is too crude to assign reliable centipawn values. While CAPS is great in theory, it requires more computing power than chess.com (or your home computer) puts into it.

"the engine search depth is much too shallow"

Ok, that should explain it. I knew something wasn't right.

i think caps is a better rating system

Caps is a better rating system than what?

then chess.com accuracy

I thought the assertion was chess.com's accuracy is based on the caps system. How are they different? What makes the caps better than the chess.com accuracy?