Chess.com computer analysis/report is very unreliable

It's not the first time that happens.

At the begining I trusted chess.com computer analysis/report, because I wasn't paying careful attention. But sometimes when I check out the analysis I find ridiculous results. For example, in the next analysis I show what I mean:

According to the analysis, 8. Nc3 is a mistake I did, because I missed the chance to get 2 whole points advantage. When you look at the reason I could get those two points with the suggested move 8. c5, you can realize that you only get that advantage if the opponent suicides his knight with the move 9. ...Nxd4 after I make a previous mistake 9.gxf3 instead of 9.Bxf3 (the suggested analysis move exposes my king with a weakened pawn structure).

So, why the analysis says my opponent would give me his knight as a gift, just because he would like  to lose in that random moment?

And because of that hypothetical gift, it's a mistake if I don't play that line? And that line is not a forced line, so it can't be a mistake.

Here there is the screenshot of the analysis:

That's the game. I suppose a new analysis could give different results.

I'm running Chrome browser on a good computer:

Graphics card: NVIDIA GeForce GTX 1050

I like it a lot the analysis interface, stats... etc in chess.com. It's sad that it's providing this ridiculous results, because that makes all statistics about the analysis unreliable too.

Why don't they use stockfish 11 (freeware) for the analysis? Maybe that way it could be more reliable.

I didn't read what you said but it's only depth 20 so the analysis could be unreliable. If you want to seriously analyze a game don't do it chess.com unless you are diamond member maybe.

Always trust the computer :-) The reason you are winning with c5 is that his knight on c6 is lost (e.g. after Qa4 and Bb5). And thats why you take on f3 with the pawn (to still have Bb5! 

I believe it is a matter of depth, not the brand of the engine.

I agree with toldsted, 8.c5 is winning big, threatening a pin on the knight with Bb5.  8. Qa4 is also interesting creating the pin straightaway.  I think the engine got it right this time.

I just analysed the position with Stockfish 11 in ChessX GUI (Both free software).

The result is really cool: chess.com engine is correct, the sacrifice of the knight is necessary anyways, but the line is so deep and full of surprises that it's really difficult to see. It's counter-intuitive.

Even looking now at the position at the end of this variation I can't see the 2 points gain! the rook is more valuable than a Bishop and I don't see a positional decisive advantage 2 points worth.

I'm very surprised with this line full of unexpected moves.

Here is the analysis of that position:

Ok, I retract for what I said about chess.com analysis  (for now, hehehe). They just represent what the engine thought, but what the engine thinks is not their responsibility, since they are not the engine designers. So we have to trust their analysis (If we trust the powerful engine).

I should have analysed it first with this software before posting this. We all have the right to a second opinion (in medicine and chess, heheheh). Anyways it's a cool variation to enjoy for everyone  

Thank you for your comments and opinion, and it's still under question if that position is worth 2 points. I think it's because 18.Qb7 Re8 and the Queen would take those two pawns (a and c). Two pawns worth two points, but a rook is 5 points and a bishop 3 points. So those two extra points must be because the white a and c pawns would be passed pawns... I suppose.

Next is the the game in pgn with variation included (with the Stockfish 11 in ChessX analysed variation, not the chess.com analysed variation):

As a conclusion, in this last match the analysis of that particular move was ok.

However, comparing in other matches I have carefully checked that the analysis given by chess.com compared with the analysis given by other software (like Lucas chess with stockfish 11 engine, both free software), chess.com only will detect blunders and big mistakes, but there are good moves and other mistakes that are missing by chess.com analysis. But that's what a fast analysis is.

If we want a real good analysis, we have to do it with other free software (Arena, Lucas chess, Scid vs Pc, Winboard, etc. with stockfish 11 engine), because you can adjust how deep you want the analysis, so you can select a decent depth.

Lucas chess is the one I recommend, because it shows plenty of statistics, even those not shown by chess.com, is highly configurable, it comes with lots of features and it's free!

I don't know how is the analysis for chess.com premium account. But it will not be better than free software, at most it can only be the same because there are no other engine stronger than stockfish right now.