Sometimes analysis is not accurate

In this game, I have forked the opponent's rook and queen. The game analyzer however says, I will lose a bishop, which is ludicrous.

Also, the analyzer often discounts tactical sacrifices and term them blunder.

Another game, the analyzer fails to recognize a development focusing towards checkmate.

You've mis-evaluated- your fork fails.

1. He's on mobile so the engine is at its weakest with the inferior hardware compared to PC

2. Chess.com limits the analysis engine's strength setting for non-diamond users

3. The problem here isn't even the engine, it's the game review's interpretation of the engine results, the system that writes out the move descriptions "This move choice loses a bishop"

Thanks for the insights. That's helpful. It is fair enough to agree that the engine is robust from the POV of a beginner or novice player. I probably had to have a better sample here, as I found some of the suggestions to be not so accurate. I think as Rob said, it might be because I am on my mobile phone.

You're right. The analysis is wrong on very rare occasions. This just happened to not be one of them, due to a rather obtuse counterplay.

That is because those are blunders.

If your "tactical sacrifices" work, they only work because your opponent defends badly. The engine is extremely strong in tactics.

I don't have the right examples here, will post one when I get it. I concur that many opponents don't defend properly, but sometimes you sacrifice your queen to achieve checkmate, in a move or two. But the engine suggested alternate moves that offers no benefits whichever way you look at it.

Whenever you think that the engine is wrong, check the line it offers, plus try checking your idea and see if the engine can refute it.

Generally you will find that it is not wrong, as in the example you've posted above and @ItsTwoDuece provided the explanation why the engine is correct there, and how the fork fails.

There is almost zero chance that engine will not see a forced checkmate in 2 or 3. If it says it is a blunder in such a case, it is basically almost certainly a blunder. Bear in mind that engines are stronger than the best players in the world.

As for a phone / weaker engine (or with smaller depth of the engine), it might get things wrong in a sense that it evaluates a certain move as +2.3, and after the move is made it evaluates it as like +1.2 instead (because the engine gets to reevaluate a lot of moves further down the line, after the move is made), but not when there is a checkmate in 2 or 3.

That is too easy for it to evaluate, and the engine will not get those wrong.

There is absolutely no way that the engine is wrong and you are right in this type of tactical position. It's completely impossible. The only explanation is that you seriously miscalculated something and there is no mate.

I am not trying to say rhat I am right and the engine is wrong. That's not what I meant. I am new here, I just wanted to understand whether the engine also considers tactical sacrifices or not, as such moves are not suggested when you analyze after the game, and they are termed blunders or mistakes.

It's hard to know what you mean by "tactical sacrifices". The answer is yes, the engine calculates sacrifices too. It considers every move. When the engine says that it's a blunder, the sacrifice doesn't work.

The engine does consider sacrifices- if the best move is a sacrifice, it will reflect that and deem it best or even great or brilliant. If the engine deems a sacrifice a mistake, that means there was essentially certainly a more winning alternative.

That being said, this doesn't mean that inaccurate sacrifices are inherently bad, in human terms sometimes they are the most practical option (because they might exploit your opponents human psychology or time for example), but they are not the objectively best option.

If the "tactical sacrifice" you are thinking of is not mentioned when you review the game, that means it's objectively not one of the best moves. A decisively winning tactical sacrifice will be mentioned. For example, here is an old game of mine. If you put it through the review, it deems multiple sacrifices as best or brilliant. You'll notice that sacrifices 14... Bxh3, 15... Bxg2, and 20... Rxd4+ are all deemed brilliant, while the sacrifice 19... Rxe4+ is inaccurate. This is because the brilliant three were all the best move, while the inaccuracy missed an immediate checkmating opportunity:

That is game review. These comments are in many cases not very useful. Basically you have to figure out why the engine shifts its evaluation. It is better to use analysis instead of game review, and check the lines engine suggests, or try some move yourself.

Here is why the move is bad. White can play Ng6+. This is a check because queen checks the king. On the next move white can take the rook with the knight.

So Ng6+, black has to deal with the check so for instance Be7, and then Nxh8.

There are other good moves, but that one is probably the simplest.

You're right. I am so embarassed right now. I am playing only bullet games right now, and my mind has become so dumb and unidirectional (lol). Why didn't I think of that!