And to clarify, the engine doesn't have to search every possible move. Finding a mate in #26 sounds impossible, but it's not. Forcing moves makes it possible to prune the search tree, but it is of course much harder to find a guaranteed mate than just a very likely mate.
An engine says '#x' and then x gets bigger when you play recommended moves? If you can reproduce that, please provide FEN and name of engine.
@schlawg said in #12:
> x gets bigger when you play recommended moves
No, not so: it’ve seen as it got bigger while waiting for Stockfish to calculate the same position more deeply. I didn’t move any pieces.
> x gets bigger when you play recommended moves
No, not so: it’ve seen as it got bigger while waiting for Stockfish to calculate the same position more deeply. I didn’t move any pieces.
@schlawg said in #12:
> An engine says '#x' and then x gets bigger when you play recommended moves? If you can reproduce that, please provide FEN and name of engine.
I think what he means is, as previously mentioned before, that at for example depth 30, it says a mate in 23... Then at depth 50, it finds a better defense for black so it says mate in 26.. Calculating in deeper depth is more accurate and it can change the eval of the position
> An engine says '#x' and then x gets bigger when you play recommended moves? If you can reproduce that, please provide FEN and name of engine.
I think what he means is, as previously mentioned before, that at for example depth 30, it says a mate in 23... Then at depth 50, it finds a better defense for black so it says mate in 26.. Calculating in deeper depth is more accurate and it can change the eval of the position
Is there a FEN for this position?
The stockfish devs will likely be interested if you've found a false "forced mate" guarantee that is revoked or extended upon deeper analysis. That's not supposed to happen.
@schlawg said in #16:
> The stockfish devs will likely be interested if you've found a false "forced mate" guarantee that is revoked or extended upon deeper analysis. That's not supposed to happen.
True!, but I believe there is FEN for the Position. And when you said, "That's not supposed to happen." It can, but only in glitches or malfunctions. So it is kinda possible.
> The stockfish devs will likely be interested if you've found a false "forced mate" guarantee that is revoked or extended upon deeper analysis. That's not supposed to happen.
True!, but I believe there is FEN for the Position. And when you said, "That's not supposed to happen." It can, but only in glitches or malfunctions. So it is kinda possible.
@HeckerIGH said in #17:
> True!, but I believe there is FEN for the Position.
If someone will provide it, I can look into it.
> True!, but I believe there is FEN for the Position.
If someone will provide it, I can look into it.
@schlawg said in #16:
> The stockfish devs will likely be interested if you've found a false "forced mate" guarantee that is revoked or extended upon deeper analysis. That's not supposed to happen.
FWIW, I've experienced what OP describes on multiple occasions and was similarly confused. Not sure if the forced mate was false but definitely seen it stop appearing or increase in # of moves. No memory of which games but will pass on next time I see it occur.
> The stockfish devs will likely be interested if you've found a false "forced mate" guarantee that is revoked or extended upon deeper analysis. That's not supposed to happen.
FWIW, I've experienced what OP describes on multiple occasions and was similarly confused. Not sure if the forced mate was false but definitely seen it stop appearing or increase in # of moves. No memory of which games but will pass on next time I see it occur.
@schlawg
The FEN is this one: “8/8/7p/pp1p2p1/P3k1P1/1PP4P/4K3/8 b - - 0 1”. It’s a position of the puzzle #3GiCS.
I have recorded Analysis Board’s local Stockfish going down to depth 63, and the anomaly in question occurs several times through the journey, so at least I can be sure that I hadn’t just imagined it. But apparently Stockfish is not-quite-deterministic—it goes a little bit different routes each time.
The settings I used are as such, just in case it matters:
Engine: Stockfish 16 NNUE · 40 MB
Search time: infinite
Lines: 3
Threads: 2
Memory: 512 MB
I surely would like to share the recording—but if only I knew where to upload it without going into hassle; that is neither YouTube nor Vimeo.
The FEN is this one: “8/8/7p/pp1p2p1/P3k1P1/1PP4P/4K3/8 b - - 0 1”. It’s a position of the puzzle #3GiCS.
I have recorded Analysis Board’s local Stockfish going down to depth 63, and the anomaly in question occurs several times through the journey, so at least I can be sure that I hadn’t just imagined it. But apparently Stockfish is not-quite-deterministic—it goes a little bit different routes each time.
The settings I used are as such, just in case it matters:
Engine: Stockfish 16 NNUE · 40 MB
Search time: infinite
Lines: 3
Threads: 2
Memory: 512 MB
I surely would like to share the recording—but if only I knew where to upload it without going into hassle; that is neither YouTube nor Vimeo.
This topic has been archived and can no longer be replied to.