91% Accuracy? Are You Mocking Me?

Oh, someone analyzed it at a higher depth. I'm down to 89.3%

No longer an 'A' grade

I posted a thread citing a specific game demonstrating how the total of best/excellent/good moves given in the Report can vary substantially from the totals given in the analysis.

https://www.chess.com/forum/view/site-feedback/this-issue-needs-fixing

My suggestion is no amount of fixes will ever make the tool viable. A quick blunder check tool is a good idea in principle- but quite impractical to assign a numerical score to the entire game. 

I'm not exactly sure what 2 things you are comparing. Screenshot the 2 parts so I can be sure. I checked that game of yours and for me it works fine. 

This is the report at 22 depth. What do you mean the analysis? Do you click on analysis then count the moves?

Yes. After each and every move the analysis reports whether the move was best/ excellent or good. Add up the totals of best moves in analysis, excellent moves in analysis and same for good moves. The Totals vary substantially from the totals given in “the Report” as seen in the screen shot.

Wow ! Holy smokes !!!

I just see the report is different than when ran before !!! Giving different totals... LOL

You all will think I’m nuts... but the Current Report is different than the one given after the game. I ran it twice and double checked everything. 
I submitted a ticket pointing out the inconsistency- received a reply from Tech agreeing - saying they were working on it. Looks like they did !

Hmm, interesting.

I counted some of it and it adds up. Bit annoying that it shows for both players making it a chore to count the best moves, which I didn't.

The original report showed 

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.

Should have taken a screen shot 

The “analysis” gives 3 book, 14 best, 9 excellent, 1 good, 1 mistake.

So the Game report is quite different today than when 1st employed. Obviously changes have been made.

What is the formula that chess.com uses to show you the accuracy?

I recently saw a postgame report on a competitor's site (presumably similar to or the same as Chess.Com's) which marked black's third move in the mainline King's Indian Defense (1. d4 Nf6 2. c4 g6 3. Nc3 Bg7) as an "inaccuracy," preferring the mainline Gruenfeld Defense third move of 3...d5.

Considering that the King's Indian Defense is played over twice as often as and wins a higher percentage of the time for black in comparison to the Gruenfeld Defense (according to a well known opening database) makes that "inaccuracy" mark a bit harsh to me and why I cite this as an example as to the unreliability of these postgame reports.

The tool is constantly being updated by programmers - assigning new “values” to opening moves. Gambits previously were always assigned “mistake” . As just pointed out 2nd /3rd opening move choices often downgraded- all depended on a % of wins seen in data bases.  A very selective and subjective report is seen - all based on human programming. 
The tool is a quick blunder check. Ok ... so far so good. But assigning an Accuracy score completely mucks up it’s purpose and validity.

The postgame report I cited didn't prefer the Gruenfeld Defense to the King's Indian Defense because the Gruenfeld wins a higher percentage of the time for black, it doesn't, but presumably because it evaluates the Gruenfeld slightly higher, which the database does.

Of course these opening database win percentages may differ from what would be expected from the evaluation for many reasons.  That would be a topic for a completely different forum though.

you guys talk weird mouth

What do you mean analysis gives those amounts? Is the report you mention the same as I screenshotted and then the others you counted yourself or where are those numbers displayed?

As explained- the tool has been updated since. It now gives a different result and totals than previously shown. It is constantly being “fixed.”