- Blind mode tutorial
lichess.org
Donate

Stockfish is weakened?

#31

@celinofj said in #21:

I've made a question: Is stockfish (lvl 8) on Lichess weaker than the regular engine?

You can restrict Stockfish with the parameters "UCI_Elo" and "UCI_LimitStrength" and I think this is what has been done.

@celinofj said in #21: > I've made a question: Is stockfish (lvl 8) on Lichess weaker than the regular engine? You can restrict Stockfish with the parameters "UCI_Elo" and "UCI_LimitStrength" and I think this is what has been done.
#32

@jomega said in #29:

It is weakened. PM me if you'd like pointers to details. However, understanding how it works eventually requires reading code, unless you want to take a stranger's word for how it works.

Didn't see your post when I posted mine. So as I understand it "UCI_Elo" and "UCI_LimitStrength" were not used, were they?

@jomega said in #29: > It is weakened. PM me if you'd like pointers to details. However, understanding how it works eventually requires reading code, unless you want to take a stranger's word for how it works. Didn't see your post when I posted mine. So as I understand it "UCI_Elo" and "UCI_LimitStrength" were not used, were they?
#33

@Katzenschinken said in #32:

Didn't see your post when I posted mine. So as I understand it "UCI_Elo" and "UCI_LimitStrength" were not used, were they?

I believe that is currently correct. Lichess seems to be using the UCI option 'Skill Level' and other means to weaken Stockfish.
See these and further down in that code where the variables LVL_MOVETIMES and LVL_DEPTHS are used to set things up for the "go" command.
https://github.com/niklasf/fishnet/blob/53e218bd184554265748513f85d19e601f18c310/fishnet.py#L117-L119
https://github.com/niklasf/fishnet/blob/53e218bd184554265748513f85d19e601f18c310/fishnet.py#L818

So I think the 1-8 'level' is used to index into these:
LVL_SKILL = [0, 3, 6, 10, 14, 16, 18, 20]
LVL_MOVETIMES = [50, 100, 150, 200, 300, 400, 500, 1000]
LVL_DEPTHS = [1, 1, 2, 3, 5, 8, 13, 22]

Edit: This is an old blob. Looking for the current code.
The new code seems to work in similar fashion, though now the SF range of levels extends to negative values!
https://github.com/niklasf/fishnet/blob/master/src/api.rs#L221-L270
The move times remain the same as above. The skill is mapped as [-9, -5, -1, 3, 7, 11, 16, 20].
The depth is now mapped as [5, 5, 5, 5, 5, 8, 13, 22].

@Katzenschinken said in #32: > Didn't see your post when I posted mine. So as I understand it "UCI_Elo" and "UCI_LimitStrength" were not used, were they? I believe that is currently correct. Lichess seems to be using the UCI option 'Skill Level' and other means to weaken Stockfish. See these and further down in that code where the variables LVL_MOVETIMES and LVL_DEPTHS are used to set things up for the "go" command. https://github.com/niklasf/fishnet/blob/53e218bd184554265748513f85d19e601f18c310/fishnet.py#L117-L119 https://github.com/niklasf/fishnet/blob/53e218bd184554265748513f85d19e601f18c310/fishnet.py#L818 So I think the 1-8 'level' is used to index into these: LVL_SKILL = [0, 3, 6, 10, 14, 16, 18, 20] LVL_MOVETIMES = [50, 100, 150, 200, 300, 400, 500, 1000] LVL_DEPTHS = [1, 1, 2, 3, 5, 8, 13, 22] Edit: This is an old blob. Looking for the current code. The new code seems to work in similar fashion, though now the SF range of levels extends to negative values! https://github.com/niklasf/fishnet/blob/master/src/api.rs#L221-L270 The move times remain the same as above. The skill is mapped as [-9, -5, -1, 3, 7, 11, 16, 20]. The depth is now mapped as [5, 5, 5, 5, 5, 8, 13, 22].
#34

@celinofj said in #21:

?

I don't understand you, guys...

It's a lichess forum. Nobody understands anything.

@celinofj said in #21: > ? > > I don't understand you, guys... > It's a lichess forum. Nobody understands anything.
#35

@celinof yes stockfish level 8 is much weaker than the original version which you can get in the analysis because it calculates less in depth with each move from what I was told in the forum, but one thing is certain is that the level 8 is much less strong than that of the analysis because I made it face a few months ago, if nothing has not changed yes he is weakened

there has never been a level 9 or 10 so I don't know since when it was weakened

or if it has been less strong the whole time

@celinof yes stockfish level 8 is much weaker than the original version which you can get in the analysis because it calculates less in depth with each move from what I was told in the forum, but one thing is certain is that the level 8 is much less strong than that of the analysis because I made it face a few months ago, if nothing has not changed yes he is weakened there has never been a level 9 or 10 so I don't know since when it was weakened or if it has been less strong the whole time
#37

When you are playing Stockfish levels 1 - 8 on Lichess, you may be actually playing Fairy Stockfish - a version of Stockfish that can play the chess variants on Lichess. If you play standard chess, then you are playing Stockfish 14 + NNUE, with parameters passed via the UCI interface to make it weak.

In particular, what people are calling Stockfish level 8, is, for standard chess, Stockfish 14 + NNUE, with the "Skill Level" parameter set to 20, the "movetime" of the "go" UCI command set to 1000 ms, and the "depth" of the "go" command set to 22.

See my post in #33 and for the UCI protocol see
http://wbec-ridderkerk.nl/html/UCIProtocol.html

When you are playing Stockfish levels 1 - 8 on Lichess, you may be actually playing Fairy Stockfish - a version of Stockfish that can play the chess variants on Lichess. If you play standard chess, then you are playing Stockfish 14 + NNUE, with parameters passed via the UCI interface to make it weak. In particular, what people are calling Stockfish level 8, is, for standard chess, Stockfish 14 + NNUE, with the "Skill Level" parameter set to 20, the "movetime" of the "go" UCI command set to 1000 ms, and the "depth" of the "go" command set to 22. See my post in #33 and for the UCI protocol see http://wbec-ridderkerk.nl/html/UCIProtocol.html
#38

@jomega Thanks for the finds in the source. That was exactly what I was looking for.

@jomega Thanks for the finds in the source. That was exactly what I was looking for.

This topic has been archived and can no longer be replied to.