BBN vs R+Ps ending and Rybka derivatives
Posted: Thu Jan 27, 2011 4:38 am
Not sure which subforum to put this in, but as "derivatives" seem of interest, it seemed technical, and so I put it here.
Quoting http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=20683
My impression is that the White pawns will eventually fall (maybe after the horizon), and then BBN vs R is a general win.
EDIT: Houdini 1.5 also shows a decent Black edge (no GTBs used). Only Rybka seems to give 0.00.
EDIT: TBs don't matter much, as all the pawns have to fall, and BBN vs R is going to be a strong Black advantage in any case. The real issue is how passed pawns are weighted.
EDIT: I checked an early IvanHoe, with the same result (-0.8 after a minute or so). I checked RobboLito, again with essentially the same result. So my observations/conclusion contrasts with that of the poster -- Rybka is the only engine I have that gives 0.00, and none of the derivatives I tested yielded this.
Quoting http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=20683
3k4/b3n1R1/P1b5/2P5/1P6/8/8/1K6 b - - 0 45Other than the fact that IvanHoe (latest version) with no TBs doesn't give 0.00 after a few minutes, but rather -1.2 or so (already -0.8 at the early depths), while Rybka (again no TBs) starts with a positive score for White and then after a few seconds switches to 0.00, I haven't checked this much.I even installed the necessary kbbnkr 6-man tablebases, and Rybka still claimed draw, as do its [her?] derivatives, and as does (unfortunately) Naum. All other post-Crafty engines correctly identify that this is completely won for black, something that IDeA eventually agrees with after a very long time of interaction (though still not traced back to move 45, which it still thinks is a draw).
My impression is that the White pawns will eventually fall (maybe after the horizon), and then BBN vs R is a general win.
EDIT: Houdini 1.5 also shows a decent Black edge (no GTBs used). Only Rybka seems to give 0.00.
EDIT: TBs don't matter much, as all the pawns have to fall, and BBN vs R is going to be a strong Black advantage in any case. The real issue is how passed pawns are weighted.
EDIT: I checked an early IvanHoe, with the same result (-0.8 after a minute or so). I checked RobboLito, again with essentially the same result. So my observations/conclusion contrasts with that of the poster -- Rybka is the only engine I have that gives 0.00, and none of the derivatives I tested yielded this.
