Author: Robert Hyatt
Date: 11:29:18 01/02/02
Go up one level in this thread
On January 02, 2002 at 11:40:54, Uri Blass wrote: >On January 02, 2002 at 11:26:06, Robert Hyatt wrote: > >>On January 02, 2002 at 11:19:52, Uri Blass wrote: >> >>>[D]r1b4k/ppp2Bb1/6Pp/3pP3/1qnP1p1Q/8/PPP3P1/1K1R3R w - - bm Qd8+; >>> >>>It is WAC96 >>> >>>Qd8+ is a forced mate for white but b3 is also enough to win the game. >>> >>>I saw cases when stupid moves are considered as solutions for WAC only because >>>they are winning so in this case b3 should be considered as an alternative >>>solution: >>> >>>after b3 I gave yace to learn that Na3+ Kc1 is bad for black and it give the >>>folllowing analysis: >>>New position >>>r1b4k/ppp2Bb1/6Pp/3pP3/1qnP1p1Q/1P6/P1P3P1/1K1R3R b - - 0 1 >>> >>>Analysis by Yace 0.99.56: >>> >>>1...Qxb3+ 2.axb3 >>> +- (9.15) Depth: 1 00:00:00 >>>1...Nxe5 2.dxe5 >>> +- (2.61) Depth: 1 00:00:00 >>>1...f3 2.gxf3 >>> +- (1.57) Depth: 1 00:00:00 >>>1...Rb8 2.Bxd5 >>> ± (1.34) Depth: 1 00:00:00 >>>1...Bf5 2.Bxd5 >>> ± (1.12) Depth: 1 00:00:00 >>>1...c6 2.Qxf4 >>> ± (1.04) Depth: 1 00:00:00 >>>1...Ne3 >>> ² (0.26) Depth: 1 00:00:00 >>>1...Ne3 2.Qd8+ Qf8 >>> ² (0.48) Depth: 2 00:00:00 >>>1...Ne3 2.Qd8+ Qf8 3.Qxf8+ Bxf8 4.Rdg1 >>> = (0.07) Depth: 3 00:00:00 >>>1...Ne3 2.Qxf4 Bxe5 3.Rxh6+ Kg7 4.Qxe5+ Kf8 >>> ² (0.47) Depth: 4 00:00:00 >>>1...Ne3 2.Qxf4 Nf5 3.Bxd5 >>> +- (2.27) Depth: 4 00:00:00 >>>1...Bf5 2.Bxd5 Bxc2+ 3.Kc1 >>> +- (2.26) Depth: 4 00:00:00 >>>1...Bf5 2.Qxf4 Bd7 3.Bxd5 >>> +- (1.91) Depth: 4 00:00:00 >>>1...f3 2.Qd8+ Qf8 3.Qxf8+ Bxf8 4.bxc4 fxg2 5.Rdf1 >>> +- (1.90) Depth: 4 00:00:00 >>>1...f3 2.Qd8+ Qf8 3.Qxf8+ Bxf8 4.bxc4 fxg2 5.Rhg1 >>> +- (1.83) Depth: 4 00:00:00 >>>1...Nb6 2.Qd8+ Qf8 3.Rde1 >>> +- (1.82) Depth: 4 00:00:00 >>>1...Nb6 2.Qxf4 Qf8 3.Rdf1 >>> +- (1.76) Depth: 4 00:00:00 >>>1...Nb6 2.Qd8+ Qf8 3.Qxc7 Bg4 4.Rdf1 >>> +- (1.62) Depth: 5 00:00:00 49kN >>>1...Nb6 2.Qd8+ Qf8 3.Qxc7 Bg4 4.Rdf1 Qc8 5.Qxc8+ Rxc8 6.Rxf4 >>> +- (1.82) Depth: 6 00:00:00 129kN >>>1...Qf8 2.Kc1 Bf5 3.Qg5 Qa3+ 4.Kb1 >>> +- (1.81) Depth: 6 00:00:00 129kN >>>1...Qf8 2.Qxf4 Na3+ 3.Kc1 c6 4.Rdf1 Bd7 >>> +- (1.41) Depth: 6 00:00:00 129kN >>>1...Qf8 2.e6 Na3+ 3.Kc1 c6 4.e7 Bf5 5.exf8Q+ Bxf8 >>> +- (1.81) Depth: 7 00:00:01 402kN >>>1...Qf8 2.e6 Bxe6 3.Bxe6 Qa3 4.Qf6 Bxf6 5.bxc4 >>> +- (1.81) Depth: 7 00:00:01 402kN >>>1...Qf8 2.Qxf4 Na3+ 3.Kc1 b6 4.Rxh6+ Bxh6 5.Bxd5 >>> +- (1.49) Depth: 8 00:00:03 1176kN >>>1...Qf8 2.Qxf4 Na3+ 3.Kb2 Bd7 4.Rxh6+ Bxh6 5.Rh1 Qxf7 6.gxf7 Kg7 7.f8Q+ Rxf8 >>>8.Qxh6+ Kf7 >>> +- (1.89) Depth: 9 00:00:07 2235kN >>>1...Qf8 2.Qg5 Ne3 3.Rxh6+ Bxh6 4.Rh1 Ng4 5.e6 Bxe6 6.Bxe6 f3 7.Qxg4 >>> +- (2.33) Depth: 9 00:00:11 4029kN >>>1...Qf8 2.Qg5 Ne3 3.Rxh6+ Bxh6 4.Qf6+ Qg7 5.Rh1 Bh3 6.Rxh3 Ng4 7.Qxf4 Qxf7 >>>8.gxf7 Nxe5 >>> +- (2.73) Depth: 10 00:00:26 9166kN >>>1...Qf8 2.Qxf4 Bf5 3.Qxf5 Ne3 4.Rxh6+ Bxh6 5.Qf6+ Qg7 6.Rh1 Ng4 7.Qf4 c6 8.Qxg4 >>>Rf8 >>> +- (4.05) Depth: 10 00:00:44 15973kN >>>1...Qf8 2.Qxf4 Bf5 3.Qxf5 Ne3 4.Rxh6+ Bxh6 5.Qf6+ Qg7 6.Rh1 Ng4 7.Qf4 c6 8.g3 >>>Rf8 9.Qxg4 >>> +- (4.14) Depth: 11 00:01:53 41818kN >>>1...Qf8 2.Qxf4 Bf5 3.Qxf5 Ne3 4.Rxh6+ Bxh6 5.Qf6+ Qg7 6.Rh1 Ng4 7.Qf4 c6 8.Qxg4 >>>Qxf7 9.gxf7 Kh7 >>> +- (4.54) Depth: 12 00:03:42 79943kN >>>1...Qf8 2.Qxf4 Bf5 3.Qxf5 Ne3 4.Rxh6+ Bxh6 5.Qf6+ Qg7 6.Rh1 Ng4 7.Qf4 >>> +- (4.54) Depth: 12 00:04:00 87418kN >>>1...Qf8 2.Qxf4 Bf5 3.Qg5 Na3+ 4.Kb2 Nxc2 5.Rxh6+ Bxh6 6.Qf6+ Qg7 7.Rh1 Bxg6 >>>8.Be8 >>> +- (4.54) Depth: 13 00:11:00 218149kN >>> >>>(Blass, Tel-aviv 02.01.2002) >>> >>> >>>Uri >> >> >>If b3 isn't a forced mate, then I disagree. WAC is about finding the best >>move in a position. If several moves lead to the same score, then they are >>all reasonable solutions. But if one leads to +4 and one leads to a forced >>mate, the +4 is wrong, because the program simply isn't seeing deep enough >>to see the _real_ issue... > >programs can also play Qd8+ for the wrong reason so by this logic >program that found Qd8+ without the right score also did not solve it. > >Here is another example from WAC >Wac31 >[D]rb3qk1/pQ3ppp/4p3/3P4/8/1P3N2/1P3PPP/3R2K1 w - - bm Qxa8 d6 dxe6 g3 > >A program that plays Qxa8 at depth 1 does not see the real reason that it is >winning but Qxa8 is considered as a solution in Dann corbit's site. > > > >Uri There is an important difference... 1. A program can find the right move for the wrong reason... 2. It can find the wrong move for the wrong reason. In case 1, the PV can sometimes show whether a program sees everything or nothing, or whether it gets a "sniff" but can't see the final outcome. Whether you count this or not is a personal preference. Of course I prefer to find the right move for the right reason, but nobody really does such detailed analysis on the solutions so I don't worry about it except when testing new ideas... In case 2, finding a move that is clearly inferior is not a good idea, even if the move is possibly winning. There are several multiple-solution positions in WAC, but they lead to roughly the same winning (or drawing) advantage...
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.