Author: Vasik Rajlich
Date: 01:37:31 01/01/06
Go up one level in this thread
On January 01, 2006 at 02:53:42, Sune Larsson wrote: >On December 31, 2005 at 21:17:47, Joseph Ciarrochi wrote: > >>One possibility is that rybka does not have specific endgame knowledge, but its >>general chess knowledge is quite applicable to the endgame? >> >>It seems like rybka is one version away from being the best at analyzing >>endgames? or am i too optimistic? > > > Exciting future to come...Rybka is the crowned bullet master from rumours > I've heard....so relatively better in endings at short time control.... > relatively - compared to other engines. > > But I close my eyes for games in blitz tempo. Two games from the pos 1 > in the suite - 40 moves in 5 minutes - were enough. Rook ending and the > level of play was...huh... not good... > > With some more 40-games matches, using the endgame suite, we can study more > details - which engine seems to handle the rook ending best? How about > the theme "good knight vs bad bishop"? etc. Interesting. > > > To keep in mind - or on paper - the following positions are documented and > verfied *won* for white/black - and highly unbalanced - even if some engines > don't recognize it... > > > Won positions: N:o 10, 11, 12, 13, 14, 15, 17, 19 > > Draw position: N:o 16 (black draws with activity only - 1.- Rd2!) > > Slightly better for black: N:o 20 (the double rook ending by Andersson) > > Unclear position: N:o 18 (the bishop pair vs pair of knights + an extra pawn) > > > 1-9 are the NunnE positions. > > > /S > Sune / Joe, thanks for making this testsuite, it's a great idea. I started looking at some games between Fruit 2.2 and Rybka from these positions, and had to stop - it was just too ridiculous. It is not only Rybka which was being ridiculous. I will add the knowledge to Rybka piece-by-piece, maybe we can use this testsuite to measure the progress. Vas > > > > > >> >> >> >>On December 31, 2005 at 17:02:13, Jouni Uski wrote: >> >>>I played 60/5 level match from Sune's positions: >>> >>>1 Rybka 1.0 Beta 32-bit 0½½0½½½11½½½01100110½0½010½½½½1½½½111½½0 20.5/40 >>>2 Fritz 9 1½½1½½½00½½½10011001½1½101½½½½0½½½000½½1 19.5/40 >>> >>>Hmm?!?!? "No endgame knowledge". Quite modest! >>> >>>Jouni
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.