Are Game Review ratings correct?

Hi guys! I want to talk about the Game Review. A lot of people use it, and not for a long ago update happened: chess.com estimates ratings. But are the ratings correct? I want to share some examples from the games I saw (however, I won't say where I saw and the players' names as well as photos , only ratings).

Example 1:

This is the game:

Interesting game. If you didn't see the ratings, you would say this game is definitely below 2000. Some of you even below 1500. 5 mistakes for white, 4 mistakes + 1 blunder, hanging mate in 1 for black. Also, if I say that black is NM? No joke. But wait - what chess.com thinks?

Okay, 75% accuracy. What are the ratings?

I mean... 2250??? With 75% accuracy??? No no no no no.... This is...

Example 2:

This is the game:

Similar game, although ratings aren't even 800. First, what is the accuracy?

Oh wow, 84% vs 77%! And only 3 mistakes for black, and 0 for white! And no blunders for both of them! It's gotta be at least 2300.

WHAT??? HOW IN THE WORLD IT'S JUST 1200?? Okay, maybe black was worse by 300 points, but... White played 9% more accurate than in example 1, but got clean 1000 rating points lower. And you know what? Game Review simply looks to actual ratings and edits it a bit.

And also, you see what marks are next to the opening, middlegame and endgame? Or good, or best, or great. What were they in the example 1? Let's take a look... (Looks at it) Oh wow! Inaccuracies! Yes, opening is great, but only for NM. But the middlegame... 

Okay, let's look to the one more, final example:

Example 3:

This is the game:

I have no words. 1900 elo??? And this was blitz. How??? 10 blunders, 10 mistakes, 16 misses!!! Just look at it!

36% vs 43%!!! Oh wow. Will it be 100 rating?? Oh, I forgot, players were 1900 , so that will not be... 

Yep, as I said. 1150??? And blunders everywhere??? Just look at example 2. 84 accuracy, everything is at least, AT LEAST good and 1200 for winner, 900 for loser. And now what?? 1150??? With blunders everywhere??? If this game was by two 500, ratings definitely would be 100.

That means, Game Review looks at the ratings and adjust them. I suggest that it doesn't look at the ratings, and guess it blindfold. That would be better. Imagine you are 700 elo, get a perfect game with 95% accuracy (one in a life) and get only 1200. GM with 90% would get 3000+. So... yea.

Thank you so much for reading! If you have some thoughts - let me know and comment below (oh, that rhymes )