Author: Uri Blass
Date: 08:25:19 06/27/03
Go up one level in this thread
On June 27, 2003 at 09:38:16, Slater Wold wrote: >On June 27, 2003 at 01:56:32, John Merlino wrote: > >>On June 27, 2003 at 00:48:25, Slater Wold wrote: >> >>>On June 26, 2003 at 17:54:35, ALI MIRAFZALI wrote: >>> >>>>[D]3rr3/6k1/2p3P1/1p2nP1p/pP2p1P1/P1B1Nb1B/2P2K2/5R2 b >>>>In the above diagram Deeper Blue made a huge mistake by playing 39...Kh6 and >>>>lost.Commercial programs do not play this move but choose 39...h4 which draws. >>>>Amazingly no annotation to this game mentions this as the loosing mistake ;but >>>>they concentrate on moves like 34...Qxf1+ as mistakes.I Let Chessmaster 9000 >>>>at 1.8 GHZ play both sides at 999 seconds per move for both sides after 39...h4 >>>>and the result is a draw.Amazingly 39....h4! is played by the programs at very >>>>low depths also.Deeper Blue must be inferior to current Commercials for turning >>>>a Draw into a loss by playing 39....Kh6? 200 million positions per second and >>>>still it does not see 39...h4.I think Deeper Blue was at least 200 Elo points >>>>weaker than the current Commercials.HERE IS CHESSMASTER 9000 autoplay at 999sec >>>>per move at 1.8 GHZ after 39.....h4! >>>>CHESSMASTER 9000:39...h4 40.Ng2 Kf8 41.Nxh4 Bd1 42.f6 Bxc2 43.g7+ Kf7 44.Nf5 >>>>Kxf6 45.g5+ Kxg5 46.Bxe5 Rxe5 47.Rg1+ Kf6 48.g8=Q Rxg8 49.Rxg8 Rxf5+ 50.Bxf5 >>>>Kxf5 AND in this Drawn position the Computer(CM 9000) moves pieces back and >>>>forth for another 30 moves which I do not list without any advantage for any >>>>side.SO I THINK AFTER 39.....h4! IT IS A DRAW. >>> >>> >>>This is the most ridiculous thing I think I've ever read on CCC. >>> >>>DBII had h4, and and shook it in 29s to pick Kh6. Why? Because h4 is the worst >>>move possible. It loses *VERY* quickly. >>> >>>After h4, 40.Ng2 is a _HORRIBLE_ move. g5 is winning, and winning big. F6 >>>chooses it in seconds with a modest score. After 40...Kg8 41. Bg2 white is up >>>over +1.50. After 41...Kf8 42. Bxf3 the score is +2.00. >>> >>>And guess what?! DBII saw ALL of this, while your CM @ 999 seconds did not. >>> >>> >>>In the position of 39... in this game, black is lost. Period. DBII played the >>>*best* move, Kh6 with a losing score. In this position, h4 is losing. And >>>losing faster than Kh6. >> >>Agreed. It takes a little while on my (moderately busy) P4-2.4, but CM9000 >>(SS=12) does spot 40.g5 as the best response to 39...h4. It takes just over >>seven minutes.... >> >>Time Depth Score Positions Moves >>0:00 1/3 -0.35 3555 1...h4 2.Rg1 Re7 3.g5 >>0:00 1/4 -0.35 5176 1...h4 2.Rg1 Re7 3.g5 >>0:00 1/5 -0.15 12203 1...h4 2.Rg1 Re7 3.g5 Ree8 >>0:00 1/6 -0.10 40295 1...h4 2.Re1 Re7 3.Kg1 Rh8 4.Kh2 >>0:00 1/7 -0.05 90995 1...h4 2.Re1 Kg8 3.Kg1 Rd6 4.g5 >> Rdd8 >>0:01 1/8 -0.07 149730 1...h4 2.Re1 Kg8 3.Kg1 Kf8 4.Kh2 >> Kg7 5.g5 >>0:02 1/9 0.00 392941 1...h4 2.Re1 Kg8 3.Kg1 Kf8 4.Kh2 >> Kg7 5.g5 Kg8 >>0:08 1/10 0.04 1554257 1...h4 2.Bg2 Kf8 3.Rh1 Nxg4+ 4.Nxg4 >> Bxg4 5.Rxh4 Bxf5 6.Bf6 Bxg6 7.Bxd8 >> Rxd8 8.Bxe4 Bxe4 9.Rxe4 >>0:22 1/11 0.12 4249370 1...h4 2.Ng2 Kf8 3.Nxh4 Bd1 4.f6 >> e3+ 5.Kg3 e2 6.Rf5 Nd3 7.g7+ Kf7 >> 8.cxd3 Rxd3+ 9.Nf3 Rxc3 >>0:54 1/12 0.35 10491191 1...h4 2.Ng2 Kf8 3.g5 Bh5 4.Nf4 >> e3+ 5.Kg2 e2 6.Re1 Nf3 7.Rxe2 Rxe2+ >> 8.Nxe2 Nxg5 9.Nf4 Bf3+ 10.Kh2 Nxh3 >> 11.Kxh3 >>2:28 2/13 0.38 29059430 1...h4 2.Ng2 Bd1 3.Nf4 Kg8 4.Ne6 >> Rxe6 5.fxe6 Nxg6 6.Bf6 Rd6 7.Ke1 >> Bxc2 8.e7 Rd1+ 9.Ke2 >>7:12 3/14 0.74 86134041 1...h4 2.g5 Kf8 3.Bg2 Bxg2 4.Kxg2 >> c5 5.bxc5 Rd7 6.Bxe5 Rxe5 7.Kh3 >> Rxc5 8.Kxh4 Rc3 >> >>jm > >Both CM & Movie like g5 Kf8 Bg2 Bxg2; DBII did not. I bet if you looked real >deep, 3...Bxg2 would fail low, and the score would triple, or more. I *think*. > >Bxg2 is an easy move, for any engine. And DBII passed it up. I am not sure if Kh6 is better than h4(both moves are probably losing) Here is some analysis by Fritz8 It likes Kh6 but changes it's mind later to h4 and I suspect that the same could happen with deeper blue. I do not have the logfile of deeper blue near me but even if it's line is better(when it considered h4) it does not prove that deeper blue saw the tactics because better line can be a result of a slightly different evaluation. New game [D]3rr3/6k1/2p3P1/1p2nP1p/pP2p1P1/P1B1Nb1B/2P2K2/5R2 b - - 0 1 Analysis by Fritz 8: 1...hxg4-- ² (0.44) Depth: 7/23 00:00:00 81kN 1...hxg4 2.Bxg4 Kf6 3.Bxf3 exf3 4.Rg1 c5 5.Ng4+ Kxf5 6.Nxe5 cxb4 ± (1.06) Depth: 7/27 00:00:00 177kN 1...Rd6! ± (1.03) Depth: 7/27 00:00:00 184kN 1...Rd6! ± (0.87) Depth: 7/27 00:00:00 193kN 1...Rd6 2.g5 Kf8 3.Bxe5 Rxe5 4.c4 Rd2+ 5.Kg3 ± (0.75) Depth: 7/27 00:00:00 200kN 1...Kf6! ± (0.72) Depth: 7/27 00:00:00 225kN 1...Kf6 2.Rg1 hxg4 3.Bxg4 Bxg4 4.Nxg4+ Kxf5 5.Nxe5 Rxe5 6.Bxe5 ² (0.62) Depth: 7/27 00:00:00 237kN 1...Kf6-- ± (0.91) Depth: 8/23 00:00:00 292kN 1...Kf6 2.Rg1 Kg7 ± (0.91) Depth: 8/26 00:00:01 396kN 1...Rd6! ± (0.87) Depth: 8/26 00:00:01 422kN 1...Rd6! ± (0.72) Depth: 8/26 00:00:01 440kN 1...Rd6-- ± (1.00) Depth: 9/20 00:00:02 633kN 1...Rd6 2.g5 Kf8 ± (1.00) Depth: 9/22 00:00:02 707kN 1...h4! ± (0.97) Depth: 9/29 00:00:03 1196kN 1...h4! ± (0.81) Depth: 9/29 00:00:03 1233kN 1...h4 2.g5 Kg8 3.Bg2 Rd6 4.Bxe5 Rd2+ 5.Kg1 Bxg2 6.Rf2 ² (0.69) Depth: 9/29 00:00:03 1283kN 1...h4 2.g5 Kg8 3.Ke1 Rd6 4.Bxe5 Rxe5 5.Bg4 Bxg4 6.Nxg4 Re7 7.Ke2 ² (0.69) Depth: 10/28 00:00:04 1593kN 1...h4 2.g5 Kg8 3.Kg1 Kf8 4.Re1 Rd6 5.Kf2 Nxg6 ± (0.72) Depth: 11/28 00:00:07 2761kN 1...h4 2.g5 Kg8 3.Bg2 Kf8 4.Bh1 Rd6 5.Bxe5 Rd2+ 6.Kg1 Bxh1 7.Bc3 Re2 ² (0.59) Depth: 12/32 00:00:15 5828kN 1...h4 2.g5 Kf8 3.Bg2 Bxg2 4.Kxg2 Nf3 5.Rh1 Re7 6.Kf2 ± (0.75) Depth: 13/34 00:00:29 11999kN 1...h4-- ± (1.03) Depth: 14/35 00:00:57 24151kN 1...h4 2.g5 Kf8 ± (1.03) Depth: 14/49 00:01:08 28771kN 1...h4-- ± (1.31) Depth: 15/39 00:02:52 74580kN 1...h4 2.g5 Kf8 ± (1.31) Depth: 15/52 00:03:11 83028kN 1...Kh6! ± (1.28) Depth: 15/52 00:07:08 189665kN 1...Kh6-- +- (1.56) Depth: 16/39 00:09:30 253728kN 1...Kh6 2.gxh5 Bxh5 +- (1.56) Depth: 16/63 00:12:44 345407kN 1...h4! +- (1.53) Depth: 16/63 00:13:08 355882kN 1...h4! ± (1.37) Depth: 16/63 00:13:32 365719kN (, eim 27.06.2003) 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.