Author: Uri Blass
Date: 23:33:24 02/17/06
Go up one level in this thread
On February 17, 2006 at 20:27:54, Ryan B. wrote: >On February 17, 2006 at 19:46:23, Uri Blass wrote: > >>Vincent claimed that rybka's evaluation is very close to fruit in the following >>messgae. >> >>http://www.talkchess.com/forums/1/message.html?487109 >> >>He wrote. >> >>"I am for example debugging a position here from a testgame from diep. >>Where Diep says here +2.3 for white, Zappa says +1.0. Rybka says 0.1 there and >>Fruit also says 0.1 there." >> >>I also analyze positions with rybka and fruit and cannot find the similiarity >> >>Here is the last one. >> >>237: Uri Blass - Rybka 1.01 Beta 13d 32-bit, 68'/40+68'/40+68'/40 >>[D]7r/5Q2/4r3/5k2/3P1n2/2P5/PP3K2/8 b - - 0 1 >> >>Analysis by Fruit 2.2.1: >> >>40...Ke4 >> ² (0.41) Depth: 1/3 00:00:00 >>40...Ke4 41.Kg3 Rg6+ 42.Kf2 >> = (0.06) Depth: 2/7 00:00:00 >>40...Ke4 41.Kg1 Rg6+ 42.Kf1 >> = (0.00) Depth: 3/11 00:00:00 >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 4/13 00:00:00 >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 5/14 00:00:00 >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 6/20 00:00:00 19kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 7/20 00:00:00 39kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 8/20 00:00:00 86kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 9/24 00:00:01 198kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 10/24 00:00:01 443kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 11/31 00:00:01 1269kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 12/37 00:00:04 3965kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rh2+ 43.Kg3 Ree2 44.Qd5+ Kh6 45.a4 Rxb2 46.Qc6 Rh5 >>47.d5 Rg5+ 48.Kf3 >> = (-0.01) Depth: 12/37 00:00:10 10093kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rh2+ 43.Kg3 Ree2 44.Qd5+ Kh6 45.a4 Rxb2 46.Qc6 Rh5 >>47.d5 Rg5+ 48.Kf3 Rf5+ 49.Kg3 >> = (-0.02) Depth: 13/37 00:00:13 12345kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Rf2+ 45.Kg3 Rc2 46.Qd5+ Kh6 >>47.a4 Rxb2 48.Kg4 Rg1+ 49.Kf3 Nh4+ 50.Ke4 Re2+ 51.Kd3 >> = (-0.03) Depth: 14/40 00:00:31 30612kN >>40...Rf6 41.Qd7+ Kg5 42.a4 Rh2+ 43.Kg3 Rg2+ 44.Kf3 Rxb2 45.Qg4+ Kh6 46.Qc8 Rb1 >>47.Qh8+ Kg5 48.Qg8+ Ng6+ 49.Ke3 Re6+ 50.Kd3 Rf1 51.a5 Rf8 52.Qg7 >> = (-0.05) Depth: 14/49 00:01:32 89769kN >>40...Rf6 41.Qd7+ Kg5 42.a4 Rh2+ 43.Kg3 Rh3+ 44.Kf2 Nd3+ 45.Kg1 Rhf3 46.Qd8 Nxb2 >>47.Kg2 Nd1 48.Qg8+ Kf4 49.Qb8+ Ke3 50.Qc7 Rg6+ 51.Kh2 Rf2+ 52.Kh3 >> = (-0.07) Depth: 15/49 00:02:34 151361kN >>40...Rf6 41.Qd7+ Kg5 42.a4 Rh2+ 43.Kg3 Rh3+ 44.Kf2 Nd3+ 45.Kg2 Re3 46.Qd8 Kg6 >>47.Qg8+ Kh5 48.Qh8+ Rh6 49.Qb8 Rg6+ 50.Kh2 Re2+ 51.Kh3 Rxb2 52.Qh8+ Kg5 >> = (-0.11) Depth: 16/49 00:05:52 346856kN >>40...Rf6 41.Qd7+ Kg5 42.a3 Rh2+ 43.Kg3 Rg2+ 44.Kf3 Rxb2 45.Qg7+ Ng6+ 46.Ke4 Re6+ >>47.Kd3 Ra6 48.d5 Ra5 49.Qg8 Kh6 50.c4 Rxa3+ 51.Ke4 Re2+ 52.Kd4 Kg5 53.Qd8+ Ne7 >> = (-0.12) Depth: 17/55 00:12:20 731360kN >> >>(, 18.02.2006) >> >>237: Uri Blass - Rybka 1.01 Beta 13d 32-bit, 68'/40+68'/40+68'/40 >>7r/5Q2/4r3/5k2/3P1n2/2P5/PP3K2/8 b - - 0 1 >> >>Analysis by Rybka 1.01 Beta 13d 32-bit: >> >>40...Ke4 >> µ (-1.25) Depth: 3 00:00:00 >>40...Ke4 >> µ (-1.15) Depth: 4 00:00:00 >>40...Ke4 41.Qb7+ Nd5 >> µ (-0.99) Depth: 5 00:00:00 >>40...Kg5 41.Qg7+ Ng6 >> µ (-1.00) Depth: 5 00:00:00 >>40...Kg5 41.Qg7+ Ng6 42.d5 >> µ (-1.01) Depth: 6 00:00:00 10kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 >> µ (-1.04) Depth: 6 00:00:00 11kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 Rf6+ 43.Kg1 >> µ (-1.04) Depth: 7 00:00:00 16kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 Rf6+ 43.Kg1 Rhf8 >> µ (-1.11) Depth: 8 00:00:00 29kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 Rf6+ 43.Kg1 Rhf8 44.Qh7+ Ke3 >> µ (-1.06) Depth: 9 00:00:00 73kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg1 >> µ (-1.23) Depth: 9 00:00:02 180kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg1 Rff2 >> µ (-1.19) Depth: 10 00:00:03 312kN >>40...Kg5 41.Qg7+ Ng6 42.Qd7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg1 Rf5 46.d5 >> -+ (-1.46) Depth: 11 00:00:12 1113kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Rf5 44.Kh2 Nh4 45.Qg7+ Kh5 46.Qh7+ Kg4 >> -+ (-1.52) Depth: 12 00:00:21 2043kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Rf5 44.Qg2+ Kf6 45.Qb7 Nf4 46.Qc8 Re2 >> -+ (-1.56) Depth: 13 00:00:38 3629kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 Rf6+ 43.Kg1 Rh5 44.a3 Rff5 45.Qg4+ Nf4 46.Kf2 Rh2+ >>47.Kf1 >> -+ (-2.41) Depth: 13 00:00:55 4927kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 Rf6+ 43.Kg1 Rh5 44.c4 Rff5 45.Qg4+ Ke3 46.Qg3+ Ke2 >>47.Qg4+ >> -+ (-4.34) Depth: 14 00:01:02 6013kN >>40...Ke4 41.Qb7+ Nd5 42.c4 Kd3 43.Qxd5 Rf8+ 44.Kg3 Rg6+ 45.Kh4 Rh8+ 46.Qh5 Rxh5+ >>47.Kxh5 >> -+ (-4.37) Depth: 15 00:01:12 7167kN >> >>(, 18.02.2006) >> >> >>People can claim that rybka's strength is the search but it is a fact that it >>has important evaluation knowledge that fruit does not have. >> >>Also shredder does not have that knowledge and needs some search to find a >>significant big score for black. >> >>237: Uri Blass - Rybka 1.01 Beta 13d 32-bit, 68'/40+68'/40+68'/40 >>7r/5Q2/4r3/5k2/3P1n2/2P5/PP3K2/8 b - - 0 1 >> >>Analysis by Shredder 9 UCI: >> >>40...Ke4 >> ² (0.67) Depth: 1/1 00:00:00 >>40...Ke4 >> ² (0.67) Depth: 1/1 00:00:00 >>40...Ke4 >> ± (0.92) Depth: 2/5 00:00:00 >>40...Ke4 41.c4 >> ± (0.92) Depth: 2/5 00:00:00 >>40...Ke4 41.c4 >> ² (0.58) Depth: 3/5 00:00:00 >>40...Ke4 41.c4 Rh2+ 42.Kg3 >> ² (0.34) Depth: 3/7 00:00:00 >>40...Ke4 41.c4 >> = (0.09) Depth: 4/8 00:00:00 >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 4/8 00:00:00 >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 5/10 00:00:00 >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 6/13 00:00:00 4kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 7/21 00:00:00 16kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 8/21 00:00:00 50kN >>40...Ke4 41.Qb7+ Kf5 >> = (0.00) Depth: 9/26 00:00:01 161kN >>40...Kg5 >> = (-0.01) Depth: 9/26 00:00:01 208kN >>40...Kg5 41.Qg7+ Ng6 42.d5 Re7 43.Qd4 Rh2+ 44.Kg3 Rh3+ 45.Kf2 Rh2+ 46.Kg3 Rh3+ >>47.Kf2 Rh2+ >> = (-0.02) Depth: 9/26 00:00:01 266kN >>40...Kg5 41.Qg7+ Ng6 42.d5 Re7 43.Qd4 Rh2+ 44.Kg3 Rh3+ 45.Kf2 Rh2+ 46.Kg3 Rh3+ >>47.Kf2 Rh2+ >> = (-0.02) Depth: 10/28 00:00:01 337kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg1 Re1+ 46.Kh2 Re2+ >>47.Kg1 Re1+ >> = (-0.02) Depth: 11/32 00:00:01 791kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg1 Re1+ 46.Kh2 Re2+ >>47.Kg1 Re1+ >> = (-0.02) Depth: 12/33 00:00:02 1296kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg1 Re1+ 46.Kh2 Re2+ >>47.Kg1 Re1+ >> = (-0.02) Depth: 13/34 00:00:03 2244kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg1 Re1+ 44.Kh2 Re2+ 45.Kg3 Re3+ >> = (-0.03) Depth: 14/37 00:00:09 5671kN >>40...Kg5 41.Qg7+ Ng6 42.Qb7 Rf8+ 43.Kg3 Re3+ 44.Kh2 Re2+ 45.Kg3 Re3+ 46.Kh2 Re2+ >>47.Kg3 Re3+ >> = (-0.03) Depth: 15/40 00:00:18 12298kN >>40...Ke4 >> = (-0.04) Depth: 15/40 00:00:29 18832kN >>40...Ke4 41.Qb7+ Nd5 42.Qg7 Rf6+ 43.Kg1 Rhf8 44.Qh7+ Ke3 45.Qh3+ Kd2 46.Qh2+ Kd3 >>47.Qh3+ >> ³ (-0.38) Depth: 15/40 00:00:43 28041kN >>40...Ke4 41.Qb7+ >> ³ (-0.63) Depth: 16/42 00:00:59 38248kN >> >>(, 18.02.2006) >> >> >>I wonder if new fruit is going to have the knowledge that black is better. >> >>Note that rybka can see big score for black even at depth 3 and even if we >>consider that depth 3 of rybka is equivalent to depth 5 of other programs not >>fruit and not shredder can see big score for black of at least 0.5 pawns even at >>depth 14. >> >>Uri > >This is a search issue. Take a look at Craftys line and score. Fruit has a >better search that Crafty in most cases but you can see what Crafty does in >search that Fruit does not in this position and quickly realize this is a search >not eval issue. > >Ryan finding the win by rybka is a search issue but finding positive small score for black like 1 pawns is about evaluation and not search. Do you really believe that rybka at depth 3 see more than fruit at depth 17? Uri
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.