the depth seems to be wrong cause the first time I use this mode, depth=18 but afterward, when I reload the site, depth=10, and the accuracy changes too.
Game Review, like Game Report, only allows one run per day. After that first run, you can get a depth 10 analysis of games but it doesn't provide information on brilliant or great moves counts or give accuracy scores.
the depth seems to be wrong cause the first time I use this mode, depth=18 but afterward, when I reload the site, depth=10, and the accuracy changes too.
Game Review, like Game Report, only allows one run per day. After that first run, you can get a depth 10 analysis of games but it doesn't provide information on brilliant or great moves counts or give accuracy scores.
^^^^^
I found it pretty useful. It's basically the original Analysis function but over-engineered (in a good way). Maybe for high ELO player it's only par. But for an idiot like me, it's handy because of the feedback I get.
Yeah, it's a shame free accounts still get only one Review a day. But I only really have one good game a day anyway so... eh. 
it is good
It is awful! - Firstly, could we get the arrow back for the alternate next move suggestion. Secondly, an on/off button for the annoying coach comments... And thirdly, where you WANT to look at a suggested line, a function that lets you click through it move by move, and back and forward, rather than playing it like a video.
Right now, it's lost a good function, and gained one function that's near meaningless, yet annoying, and another that could be good, but is limited in a way that makes it confusing.
It seems broken to me. It is marking moves as excellent such as a player moving a rook to where it can be pinned to the king and won, and when a Queen was left en prise it only marked the move as inferior (a pawn move iirc) not a blunder. I don't expect it to give very deep analysis since it is running really for a fairly short time span, and not to a great depth, but I wouldn't expect it to give harmful feedback to a player.
https://www.chess.com/analysis/game/daily/369777983?tab=review
Its verdict is that this was a close game lost by a mistake.
It seems broken to me. It is marking moves as excellent such as a player moving a rook to where it can be pinned to the king and won, and when a Queen was left en prise it only marked the move as inferior (a pawn move iirc) not a blunder. I don't expect it to give very deep analysis since it is running really for a fairly short time span, and not to a great depth, but I wouldn't expect it to give harmful feedback to a player.
https://www.chess.com/analysis/game/daily/369777983?tab=review
Its verdict is that this was a close game lost by a mistake.
Low depth is going to miss things; if I click through it will automatically re-analyze at a higher depth, so can't really see what it's going to show.
At even the lowest depths possible it should recognize the queen being dropped as a blunder.
I'm not sure what the final depth was that it was run at, but here is an example where it is running at depth 26, which appears to be the highest I can set it in the settings:
So at depth 26 it thinks Nf6 is an excellent move while dropping the knight.
At even the lowest depths possible it should recognize the queen being dropped as a blunder.
I'm not sure what the final depth was that it was run at, but here is an example where it is running at depth 26, which appears to be the highest I can set it in the settings:
So at depth 26 it thinks Nf6 is an excellent move while dropping the knight.
White is winning no matter what black does, so essentially any move by black is basically considered good. My understanding is that process is going to be tempered somewhat, in such instances, but I don't know by how much or when that change might happen.
At even the lowest depths possible it should recognize the queen being dropped as a blunder.
I'm not sure what the final depth was that it was run at, but here is an example where it is running at depth 26, which appears to be the highest I can set it in the settings:
So at depth 26 it thinks Nf6 is an excellent move while dropping the knight.
It's hard to know how to handle such moves.
Let's say I'm attacking my opponent, and many moves are winning. I choose to sacrifice an additional knight to remove my opponent's last best defender, virtually guaranteeing I'll win the game. If an engine calls this move bad, then its analysis is pointless and worth ignoring.
But there are also situations like you post, where sacrificing a knight is stupid. If the engine doesn't call this moves bad then its analysis is pointless and worth ignoring.
But not all sacrifices during an attack are good. In a very similar situation sacing a knight might make my attack more difficult.
So you can see why it's hard to have automated analysis.
the depth seems to be wrong cause the first time I use this mode, depth=18 but afterward, when I reload the site, depth=10, and the accuracy changes too.