Author: Uri Blass
Date: 03:59:58 12/06/04
Go up one level in this thread
On December 06, 2004 at 06:52:33, Uri Blass wrote: >On December 06, 2004 at 00:33:31, Anthony Cozzie wrote: > >>Every so often when Zappa and Crafty play, they come up with some really >>brilliant moves and tough positions. Somehow, when I watched this game I had an >>intense image of two boxers who just refused to give up, no matter how bloodied >>they were. You can judge for yourself. The opening was a French defense, with >>Zappa defending. Zappa had a few tense moments, but it was able to get a very >>solid position, despite the horrible French bishop. Zappa thought it was about >>equal, while Crafty liked white to the tune of about +0.6. I guess the true >>value is somewhere in between; Black should draw but I think White has all the >>winning chances here. >> >>[D]2r1r1k1/3bq2p/p5pQ/Pp1pPp2/3P4/3B1R2/2P3PP/5RK1 b - - 0 24 >> >>So they shuffled, as computers are wont to do, and Crafty was finally so happy >>with its position that it gave away the a5 pawn in order to prevent the draw. >>At this point Crafty was about +0.4, while Zappa thought it was winning by 0.7 >>or so. The shuffling continued, with Zappa trying to advance its passed a pawn >>without letting Crafty checkmate it on the kingside. >> >>[D]2r4k/7p/p3b1p1/qp1pPp2/3P1R2/2rB2R1/2P3PP/4Q1K1 w - - 0 38 >> >>Here came the first problem. Zappa had counted on Kg8, but white has a draw in >>that line. >> >>[D]2r5/6kp/4b1p1/1p1pPp2/pq1P3Q/2rB2R1/2P3PP/5RK1 b - - 0 45; bm Qf8; test >>position 1 >> >>1... Kg7-g8 2. Rg3-h3 Rc3-c7 3. c2-c3 Qb4xc3 4. Bd3xb5 Qc3-b4 5. Bb5-a6 Rc8-b8 >>6. Rf1-e1 a4-a3 >> = (1.13) Depth: 11/33 00:00:08.83 4885kN (553 KN/s, 0 splits, 0 aborts) >>1... Kg7-g8 2. Qh4-f6 Rc3-c6 3. Rg3xg6 h7xg6 4. Qf6xg6 Kg8-h8 5. Qg6-h6 Kh8-g8 >>6. Qh6-g6 >> = (0.00) Depth: 12/33 00:00:12.63 7287kN (577 KN/s, 0 splits, 0 aborts) >>1... Qb4-f8 2. Rg3-h3 h7-h5 3. Rh3-g3 Kg7-h7 4. Bd3xb5 Rc3xg3 5. Qh4xg3 Qf8-b4 >>6. Qg3-d3 a4-a3 7. Rf1-f3 >> = (0.56) Depth: 12/34 00:00:28.62 17188kN (600 KN/s, 0 splits, 0 aborts) >> >>This let Crafty win the b5 pawn, and the shuffling continued. Zappa was able to >>get its a pawn to a2, but it remained hampered by its horrible bishop. Finally, >>just when Crafty thought it had equalized, Zappa struck: >> >>[D]8/r5k1/4b1p1/1q1pPp1p/3P4/2P3P1/p4K1P/R1QB4 b - - 0 68; bm f4! test position >>2 >> >>1... f5-f4 2. g3xf4 Be6-f5 3. Bd1-c2 Bf5-g4 4. Kf2-g1 Qb5-e2 5. Bc2-b3 Bg4-e6 6. >>f4-f5 g6xf5 7. Qc1-g5 Kg7-h7 >> = (1.14) Depth: 13/36 00:00:38.98 29001kN (744 KN/s, 0 splits, 0 aborts) >> >>After some complicated tactics, Zappa emerged with a Rook versus Crafty's 3 >>passed pawns. When I was just about to celebrate a pretty win, Zappa threw the >>game away with Rf4??. This is a really hard position for Zappa, because after >>Rf4 d5 it has a Rook and Bishop vs 2 connected passers, whereas after Bf5 it has >>a Rook vs 3 connected passers :) Crafty gets it at 8 ply or so though. >> >>[D]8/7k/4Prp1/7p/2BP4/2P5/7P/1b4K1 b - - 0 82; bm Bf5!; test position 3 >> >>After Rf4 Crafty was able to force the draw, although Zappa didn't realize it >>for another 10 moves. The final position: > > >Not a good test position. > >It is possible to find Bf5 for the wrong reason. >Movei also needs 8 plies to find it but needs depth 13 to find Crafty's d5(still >without a draw score). > >Note also that I expect zappa to find Bf5 after more time simply because it wins >material and from computer point of view it should not be. > >Movei after some time has fail high on Bf5 > >I am also not sure if d5 force a draw and movei does not agree with capturing >the bishop after d5 and suggests the line d5 Bf5 d6 Re4 > >At this moment of posting I see another fail high for d5 so maybe not capturing >the bishop also does not help. >I am surprised that zappa fail here relative to movei because I expect movei to >be weak in endgame considering the fact that it does not use hash for pruning. > >Uri It seems that the fail high was a wrong fail high. I also checked the position after Rxc4 and I find that movei can see a draw score after a minute of search and even has a positive score of +0.08 for white at depth 14. Maybe Crafty even missed a win but I will need more reliable program like yace to check it. Note that the version of movei that I used of movei is the public version and the cahnges that I did from it until now are only cosmetic changes like allowing more personalities and no change in the evaluation. 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.