Strange analysis / engine failure?

Hey friends : )
I'm analysing my last bullet game with the lichess built-in feature but something doesn't look entirely right to me:



since 18.Bxh7 seems killing, why does the engine keep telling 17...Nxh5 is still +0.5?

If you let it compute, it returns the correct value. (To me it happens with depth 27)

Stockfish at depth 29 says its +7.3. Give it time to think, and the engine will display the right evaluation.

#2 and #3
They are talking about the Server side computer analysis which shows Nxh5 as 0.5 when it is completely winning .

Computers aren't perfect. Though they do get better with depth.

If you think there is a misevaluation, either the computer hasn't gone deep enough to see what you see, or you're probably missing something.

@XxTDSxX and @GreatBigTiger
I'm indeed talking about the lichess server analysis tool, and also @XxTDSxX ciao bello! ;)

@CheckingTruth
yeah you got it

@danegraphics
true, but still... I mean: i play that shitty combo a lot and the server analysis usually spots it. Wondering why this time it fails

That game is more than slightly reminiscent of Lasker-Bauer 1889.

@MrPushwood
I see you're a man of culture ;)
yeah I play that double B sac since I found out that little Lasker gem <3

@Goet
Yes, that's a Stockfish bug.
I noticed it in a game at Lichess, and since then seen this bug in around 5 more games.
Reason being it takes a certain depth at which the computer yields it's analysis when one requests for but sometimes a move is analysed at a lower depth which I suppose may be the cause of all such thing.

#2 and #3
From my side, at depth 14 with 14 knodes/s, it showed me evaluation of +6.7 after Black played Nxh5.