Visualisation of top 3 computer lines • page 1/1 • Lichess Feedback • lichess.org

I may just be overlooking a button which does this, but I thought I'd ask.

Is there a way to see the top 3 (or N) lines a computer suggests?

A similar feature is on chess.com by default (and can be changed via 'Self Analysis Lines' setting)

I may just be overlooking a button which does this, but I thought I'd ask. Is there a way to see the top 3 (or N) lines a computer suggests? A similar feature is on chess.com by default (and can be changed via 'Self Analysis Lines' setting)

#2

jomega

#3

^^
Be aware though that Lichess chops off those variations from what is actually being returned from Stockfish. Usually you see a depth of 22 or 23 half-moves (depends on browser etc) at the top of the display unless the position has already been analyzed further and saved in the cloud. Yet if you expand the variations by clicking the down arrow to the right, you will see it chopped off at either 10 or 16 half-moves.

The upshot of that is that the evaluation that is displayed by Lichess, returned from Stockfish, is NOT the evaluation that Stockfish got for the position that Lichess is displaying as the tip (last position). Instead, it is the evaluation that Stockfish returned for the depth it was asked to do by Lichess, and that is deeper than the variation displayed. You can verify what I'm saying by running Stockfish in a command window (Windows) or shell (Unix/Linux) and giving Stockfish the same uci commands Lichess does.

I find this cutoff of the lines most annoying.

^^ Be aware though that Lichess chops off those variations from what is actually being returned from Stockfish. Usually you see a depth of 22 or 23 half-moves (depends on browser etc) at the top of the display unless the position has already been analyzed further and saved in the cloud. Yet if you expand the variations by clicking the down arrow to the right, you will see it chopped off at either 10 or 16 half-moves. The upshot of that is that the evaluation that is displayed by Lichess, returned from Stockfish, is NOT the evaluation that Stockfish got for the position that Lichess is displaying as the tip (last position). Instead, it is the evaluation that Stockfish returned for the depth it was asked to do by Lichess, and that is deeper than the variation displayed. You can verify what I'm saying by running Stockfish in a command window (Windows) or shell (Unix/Linux) and giving Stockfish the same uci commands Lichess does. I find this cutoff of the lines most annoying.

samburville

#4

Perfect thanks @Yeltcki - that's exactly what I was looking for.

Will keep in mind the cutoff with multiple lines, but for the most part this solves my query

Perfect thanks @Yeltcki - that's exactly what I was looking for. Will keep in mind the cutoff with multiple lines, but for the most part this solves my query

CM TBest

#5

@jomega said in #3:

^^
Be aware though that Lichess chops off those variations from what is actually being returned from Stockfish.

Not sure if this is really the case. Based on https://github.com/ornicar/lila/issues/2423 , Lichess shows what stockfish returns
It's possible that this issue is no longer accurate thru, given the stale status.

@jomega said in #3: > ^^ > Be aware though that Lichess chops off those variations from what is actually being returned from Stockfish. Not sure if this is really the case. Based on https://github.com/ornicar/lila/issues/2423 , Lichess shows what stockfish returns It's possible that this issue is no longer accurate thru, given the stale status.

jomega

#6

@TBest said in #5:

Not sure if this is really the case. Based on github.com/ornicar/lila/issues/2423 , Lichess shows what stockfish returns
It's possible that this issue is no longer accurate thru, given the stale status.

I'm not talking about the cases where Stockfish returns short lines, but instead the case where Stockfish returns long lines that Lichess is truncating. I can see the code in Lichess that does the truncation, and I can prove Lichess truncates by running Stockfish on the position and passing Stockfish the same uci commands.

@TBest said in #5: > Not sure if this is really the case. Based on github.com/ornicar/lila/issues/2423 , Lichess shows what stockfish returns > It's possible that this issue is no longer accurate thru, given the stale status. I'm not talking about the cases where Stockfish returns short lines, but instead the case where Stockfish returns long lines that Lichess is truncating. I can see the code in Lichess that does the truncation, and I can prove Lichess truncates by running Stockfish on the position and passing Stockfish the same uci commands.

CM TBest

#7

Then you should consider making a github issue? Maybe something can be tweaked. Sounds like you know more about it then me at least

dboing

#8

I think that as long as users don't use critical thinking in how they use engine analysis and automatically assume that because an engine is top in some tournament format against engine, that it does not have any possible bias or error, and that since it can beat any human, the score it gives need no further curiosity, no github issue will really get enough steam to actually have some change soon.

There needs to be more people willing to analysis the analysis tools being parachuted from engine competitions as post-game analysis helpers. This "chess has mysterious ways" attitude is not just about engine use, in my opinion, but it is most obvious there.

I think that as long as users don't use critical thinking in how they use engine analysis and automatically assume that because an engine is top in some tournament format against engine, that it does not have any possible bias or error, and that since it can beat any human, the score it gives need no further curiosity, no github issue will really get enough steam to actually have some change soon. There needs to be more people willing to analysis the analysis tools being parachuted from engine competitions as post-game analysis helpers. This "chess has mysterious ways" attitude is not just about engine use, in my opinion, but it is most obvious there.