You need to open the Move List window by clicking on MOVE LIST, below. Scroll down to the real game move 22. (In black letters.) Immediately thereafter, in colored letters, there appears a Computer Analysis suggested line beginning with 22...g4. Follow that colored line along to White's move, 29. Nxd4. Click on that move and look at the board position. Is there any doubt what Black's next move should be? Then look at what the 2500-rated Computer Analysis gives as Black's next logical move. Am I just reading this wrong somehow?

Yeah, that was pretty funny!  Maybe the computer takes after Chigorin...?

Wow, that's quite baffling..

I'll stick with my 2000~ rated computer analysis, thank you very much..

Yes, I have had one experience with computer analysis and it wasn't very helpful. I have also heard others with similar stories to yours.

I think you're absolutely right, cuendillar. From now on, I'll be using Chess.com's computer analysis for catching shallow-ply errors only. I think it's still very worthwhile for that, though.

But even there you have to be careful. For example, the game I showed above ended by resignation immediately after 31...Rh5+. Computer Analysis's suggested line from there is: 32. Kg6 (forced), but then 32...Qe7, a non-checking move which I consider to be inferior for that reason alone, even though it shows mate as being only one move away. The line I intended to use was: 32...e5+, with similar forcing checks all the way until mate on the 8th rank. But maybe that's just a matter of style.

I have HIARCS 12.1--supposedly rated at 2950 on my computer--for deep analysis if I want to go to that trouble. I haven't noticed this problem with HIARCS yet.

Computer is missing mate in 1?!

Yeah, that was the point of the original post. But that was many plies into the analysis. I think, as cuendillar pointed out, we just have to adjust our expectations as to Computer Analysis.

I'm shocked. simply shocked.

Wow.

The red shows what moves would have lead to a better position.