lichess.org
Donate

Unlimited, stronger game analysis on lichess.org

gamblej

My napkin testing says 512MB of hash will see you safely through 300+ million node searches before collisions begin to reduce efficiency.

16MB should be plenty around the Standard and 2s range. More hash is a much smaller benefit than more time or nodes.

My napkin testing says 512MB of hash will see you safely through 300+ million node searches before collisions begin to reduce efficiency. 16MB should be plenty around the Standard and 2s range. More hash is a much smaller benefit than more time or nodes.
gamblej

I should point out - that's based on a pretty thin interpretation of hashfull numbers from Stockfish search info response lines. Someone please correct me if I'm mistaken.

I should point out - that's based on a pretty thin interpretation of hashfull numbers from Stockfish search info response lines. Someone please correct me if I'm mistaken.
Kiyomomi

Screenshot 2026-10-17 085420.png

help me i have a problem

![Screenshot 2026-10-17 085420.png](https://image.lichess1.org/display?op=noop&path=WVahcoYiVlIf.png&sig=ce3f96d3eee4dcf8f45f37309287e81e08c2285a) help me i have a problem
gamblej

@Kiyomomi Can you give me a link to that game or study?

@Kiyomomi Can you give me a link to that game or study?
<Comment deleted by user>
Kiyomomi

@gamblej This is link to that game lichess.org/oMTZ9lhj
But there is one important piece of information that I need to mention first: for engines available on the Lichess website, such as Stockfish 19 - 94MB, everything works normally. However, the error only occurs when switching to engines that are not available on Lichess, such as Stockfish dev-20261011-31275420 (the engine I use to analyze games and report the error described above). I have tried many different games and other engines, such as Reckless 0.10.0-dev-6abcff05, but the error still follows the same pattern: if the game has n moves, then after finishing the analysis of the n(th) move, the analysis always jumps to the n+1(th) move to continue analyzing.

@gamblej This is link to that game lichess.org/oMTZ9lhj But there is one important piece of information that I need to mention first: for engines available on the Lichess website, such as Stockfish 19 - 94MB, everything works normally. However, the error only occurs when switching to engines that are not available on Lichess, such as Stockfish dev-20261011-31275420 (the engine I use to analyze games and report the error described above). I have tried many different games and other engines, such as Reckless 0.10.0-dev-6abcff05, but the error still follows the same pattern: if the game has n moves, then after finishing the analysis of the n(th) move, the analysis always jumps to the n+1(th) move to continue analyzing.
gamblej

Ok, so you've found a bug in external engine? Or rather, a local analysis bug that only happens with external engine. That's enough for me to go on. Thanks!

Ok, so you've found a bug in external engine? Or rather, a local analysis bug that only happens with external engine. That's enough for me to go on. Thanks!
gamblej

One question - which provider are you using?

One question - which provider are you using?
samwonelsie

After performing a server analysis of the following game:

https://lichess.org/NQhuAsVx/black

analysis editor says it was published by ahbanks. @gamblej what is the meaning of the word published here?

After performing a server analysis of the following game: https://lichess.org/NQhuAsVx/black analysis editor says it was published by ahbanks. @gamblej what is the meaning of the word published here?