It's literally checkmate but stockfish says white's winning

Sort:
MARattigan
MARattigan wrote:
Penguin4096 wrote:
MARattigan wrote:
Penguin4096 wrote:
MARattigan wrote:
marcellothearcane wrote:
Raphael wrote:

Wait i mean this extremely closed position that engines cannot analyze 

How can i forgot that one isolated g pawn 

This position ok 

Not the old one where i forgot the isolated g pawn 

And engine cannot analyze this 

 

Second example most likely can't be analysed because this could never be legal in a real game. Try using fairy stockfish and see what happens.

I don't think SF worries too much about the legality, but it does worry about storage.

I normally give it a 2GB hash size. 

If I give it a starting position with an extra row of pawns each on the players' third ranks it crashes. If I give it 4GB hash it runs. 

I didn't try OP's position, but it will probably run given large enough hash.

As for @Raphael's second example, SF can analyse it (with default 16MB hash), it just can't get the correct answer. That's quite normal.

oh ok

Sorry - misinformation.

Hash shortage doesn't apply if the position is already mate.

 

 

wait how did you do that

I ran SF15 from the Windows command prompt, then sent it UCI commands. Cuts out questions about what the GUI's doing.

The search from your position was successful with 16MB hash. Both chess.com and the site that dare not speak its name should receive the last three lines in the display, but what they do with it after that is nothing to do with SF.

Given that chess.com comes up with a cp evaluation, it most likely gave SF the wrong FEN or it never called SF and the number came from somewhere else. Interestingly if you try game review it does say 0-1, but it takes an awfully long time to complete for a game with 0 moves - I gave up.

Sorry again - still misinformation (and libelling the chess.com GUI).

I used SF15 in the example I posted, but what you are actually looking at is an SF11 "feature". SF15 gets it right, but SF11 doesn't. It's a programming bug in the engine, not the GUI.

SF11 thinks you should take the checking pawn on g2 with your h2 bishop. (Maybe that's how it wins so many games.)

 

Penguin4096
MARattigan wrote:
MARattigan wrote:
Penguin4096 wrote:
MARattigan wrote:
Penguin4096 wrote:
MARattigan wrote:
marcellothearcane wrote:
Raphael wrote:

Wait i mean this extremely closed position that engines cannot analyze 

How can i forgot that one isolated g pawn 

This position ok 

Not the old one where i forgot the isolated g pawn 

And engine cannot analyze this 

 

Second example most likely can't be analysed because this could never be legal in a real game. Try using fairy stockfish and see what happens.

I don't think SF worries too much about the legality, but it does worry about storage.

I normally give it a 2GB hash size. 

If I give it a starting position with an extra row of pawns each on the players' third ranks it crashes. If I give it 4GB hash it runs. 

I didn't try OP's position, but it will probably run given large enough hash.

As for @Raphael's second example, SF can analyse it (with default 16MB hash), it just can't get the correct answer. That's quite normal.

oh ok

Sorry - misinformation.

Hash shortage doesn't apply if the position is already mate.

 

 

wait how did you do that

I ran SF15 from the Windows command prompt, then sent it UCI commands. Cuts out questions about what the GUI's doing.

The search from your position was successful with 16MB hash. Both chess.com and the site that dare not speak its name should receive the last three lines in the display, but what they do with it after that is nothing to do with SF.

Given that chess.com comes up with a cp evaluation, it most likely gave SF the wrong FEN or it never called SF and the number came from somewhere else. Interestingly if you try game review it does say 0-1, but it takes an awfully long time to complete for a game with 0 moves - I gave up.

Sorry again - still misinformation (and libelling the chess.com GUI).

I used SF15 in the example I posted, but what you are actually looking at is an SF11 "feature". SF15 gets it right, but SF11 doesn't. It's a programming bug in the engine, not the GUI.

 

SF11 thinks you should take the checking pawn on g2 with your h2 bishop. (Maybe that's how it wins so many games.)

 

(when bishops move horizantally) 

oh so there are 2 stockfish engines here?

oh yeah lichess has stockfish 14+

Penguin4096
randomchessguy52 wrote:

turn sf 15 with cloud analysis maybe then it will see whats goin on?????????????????

ok

MARattigan

But how you're going to pinpoint a bug in SF11 by running SF15 which doesn't have the bug is not totally clear.

MARattigan
Penguin4096 wrote:
...

oh so there are 2 stockfish engines here?

...

Not here. Just SF11 and Komodo.

They used to have SF15, but switched to SF11 because SF15 had an inconvenient (but different) programming bug.

I'm running the examples I posted on my PC.

Penguin4096
MARattigan wrote:
Penguin4096 wrote:
...

oh so there are 2 stockfish engines here?

...

Not here. Just SF11 and Komodo.

They used to have SF15, but switched to SF11 because SF15 had an inconvenient (but different) programming bug.

I'm running the examples I posted on my PC.

oh

well Komodo 2706 isn't exactly correct either...  

MARattigan
Penguin4096 wrote:
MARattigan wrote:
Penguin4096 wrote:
...

oh so there are 2 stockfish engines here?

...

Not here. Just SF11 and Komodo.

They used to have SF15, but switched to SF11 because SF15 had an inconvenient (but different) programming bug.

I'm running the examples I posted on my PC.

oh

well Komodo 2706 isn't exactly correct either...  

Interesting. And very suspicious that the score is still +42. Maybe the GUI does also have a bug.

I don't know what Komodo 2706 is.

I have Komodo 8 and 12 which both seem to work OK (but they don't return any evaluation for positions that are already mate, just "bestmove 0000"). 

P3D0FYL3DIDDYNIKKA
Cool
Penguin4096
MARattigan wrote:
Penguin4096 wrote:
MARattigan wrote:
Penguin4096 wrote:
...

oh so there are 2 stockfish engines here?

...

Not here. Just SF11 and Komodo.

They used to have SF15, but switched to SF11 because SF15 had an inconvenient (but different) programming bug.

I'm running the examples I posted on my PC.

oh

well Komodo 2706 isn't exactly correct either...  

Interesting. And very suspicious that the score is still +42. Maybe the GUI does also have a bug.

I don't know what Komodo 2706 is.

I have Komodo 8 and 12 which both seem to work OK (but they don't return any evaluation for positions that are already mate, just "bestmove 0000"). 

the +42 down there is material, not suspicious at all.

I don't know chess.com's Komodo options says "Komodo 2706"

MARattigan
Penguin4096 wrote:
MARattigan wrote:
Penguin4096 wrote:
MARattigan wrote:
Penguin4096 wrote:
...

oh so there are 2 stockfish engines here?

...

Not here. Just SF11 and Komodo.

They used to have SF15, but switched to SF11 because SF15 had an inconvenient (but different) programming bug.

I'm running the examples I posted on my PC.

oh

well Komodo 2706 isn't exactly correct either...  

Interesting. And very suspicious that the score is still +42. Maybe the GUI does also have a bug.

I don't know what Komodo 2706 is.

I have Komodo 8 and 12 which both seem to work OK (but they don't return any evaluation for positions that are already mate, just "bestmove 0000"). 

the +42 down there is material, not suspicious at all.

I don't know chess.com's Komodo options says "Komodo 2706"

Oh thanks. I don't use the analysis much, I thought it was the engine evaluation.

Should have realised because there's no decimals.