Author: Uri Blass
Date: 13:24:56 11/04/02
Go up one level in this thread
On November 04, 2002 at 14:55:54, Sune Fischer wrote: >On November 04, 2002 at 14:13:41, Gerd Isenberg wrote: > >>On November 04, 2002 at 07:34:29, Sune Fischer wrote: >> >>>Going through the november issue of "Skakbladet", the chess magasin here in >>>Denmark, I found a few positions with some very deep moves, much too deep for >>>frenzee. >>> >>>Granted some of the positions are won even with the second best move, but for >>>the sake of beauty find the best move! :) >>> >> >>Hi Sune, >> >>IsiChessMMX results with amd xp2.1+ 256MB Hash: >> >>>r1r3k1/1pq1bppp/3pbn2/p3p3/P3P1PP/nNN2P2/KPP2Q2/2BR1B1R b - - 0 1 bm b5; >>>Motylev-Gelfand, Russia-World 2002 >> >>b5 found immediatly at depth 3. Increasing score from +24cp to +68cp at depth 6. >>FH at 11/20 with +114cp after 1:27. > >Very nice, frenzee cannot see it even at ply 12, it prefers Nc4. >I wonder if this is evaluation or tactics, seeing it at depth 3 is pretty fast I >think. Very natural move for a human btw, just tearing open the position :) Movei with less than 16 mbytes Ram needs depth 8 to fail high on b5(Nc4 get a score of only 0.01 for black). depth=8 +0.02 b7b5 Nodes: 2793031 NPS: 111990 Time: 00:00:24.94 Later: 8 27 3735 4065503 b7b5 g4g5 f6d7 f1h3 e6h3 c3d5 c7d8 a2a3 depth=9 +0.44 b7b5 f1b5 a3b5 c3b5 c7c2 f2c2 c8c2 c1g5 e6b3 a2b3 c2f2 Nodes: 7332254 NPS: 108177 Time: 00:01:07.78 depth=10 +0.31 b7b5 f1b5 a3b5 c3b5 c7c2 f2c2 c8c2 c1g5 a8d8 g5f6 g7f6 Nodes: 11649005 NPS: 107492 Time: 00:01:48.37 depth=11 +0.63 b7b5 g4g5 f6d7 f1b5 a3b5 c3b5 c7c2 f2c2 c8c2 c1e3 d7c5 e3c5 d6c5 Nodes: 39018032 NPS: 109441 Time: 00:05:56.52 depth=11 +0.63 b7b5 g4g5 f6d7 f1b5 a3b5 c3b5 c7c2 f2c2 c8c2 c1e3 d7c5 e3c5 d6c5 Nodes: 40875321 NPS: 109263 Time: 00:06:14.10 Note that movei has no king safety evaluation > >>>[D]3r4/ppp4p/1bp1k3/4pp2/8/BPP5/P2NKprP/R4R2 b - - 0 1 bm Rxd2; Gufeld-Kavalek, >>>OL 1962 (Kavalek's best game ever) >> >>not found in 5 minutes. Likes e4 with rather high score up to +380cp. > >I think the idea is to do a pawn raid, without the knight to defend white is >lost. No chance for movei if this is the idea Movei does not evaluate passed pawns. Movei fails high on e4 at depth 11 depth=11 +1.32 e5e4 d2c4 f5f4 c4b6 c7b6 f1f2 d8d2 e2d2 g2f2 d2c1 f2h2 a3f8 h2h1 c1b2 h1h2 b2a3 Nodes: 13437925 NPS: 141914 Time: 00:01:34.69 > >>>[D]2bq3k/2p4p/p1np4/7P/2BPPQp1/2p2P2/r7/2KR2R1 b - - 0 1 bm Ra3; Bagirov-Gufeld >>>1973, (Gufeld: maybe my best move ever) >> >>found after 1 second at depth 7/11, score +240cp. >>FHs after 0:17 and 0:54 with 308 resp. 369cp > >I got that on and off, eventually it stuck with Ra4. Movei likes Ra4 until depth 9 At depth 9 it finds Ra3. depth=9 +1.34 a2a3 Nodes: 4236225 NPS: 134142 Time: 00:00:31.58 depth=9 +1.60 a2a3 g1g2 c6b4 c1b1 c3c2 g2c2 b4c2 b1c2 a3f3 f4h6 c8d7 Nodes: 5703507 NPS: 133822 Time: 00:00:42.62 depth=10 +1.78 a2a3 g1g2 c6b4 c1b1 c3c2 g2c2 b4c2 b1c2 a3f3 f4h6 c8b7 d4d5 Nodes: 8279779 NPS: 132349 Time: 00:01:02.56 depth=11 +1.78 a2a3 g1g2 c6b4 c1b1 c3c2 g2c2 b4c2 b1c2 a3f3 f4h6 c8b7 d4d5 Nodes: 16368756 NPS: 133939 Time: 00:02:02.21 > >>>[D]2bq3k/2p4p/p2p4/7P/1nBPPQP1/r1p5/8/1K1R2R1 b - - 0 1 bm Be6, am c2; >>>Bagirov-Gufeld 1973 >> >>This position is also member (13) of the BT2630 testsuite. >> >>1:26 for Be6 with Be6 Bxe6 Nd3 +47cp >>FH after 1:40 +105cp and 1:52 with +163cp >>finally after 5:07 531cp with the line >>1...Bb6 2.Qc1 c2+ 3.Kb2 Bxe4 4.Kxa3 c5 5.Kb2 Qb6 6.Kc3 cxd1Q 7.Qxd1 Nd5+ >>8.Kxc4 Ne3+ 9.Kd3 Nxd1 > >Very nice. >So, how is it done? ;) This seems to be hard for movei It likes c2+ very much and only at depth 9 starts to fail low from 2.82 to 2.52 but at depth 10 the score goes up again to 2.53 depth=10 +2.53 c3c2 b1b2 c2d1q g1d1 a3h3 d1f1 c8e6 f4h6 h3h2 b2b1 e6g8 d4d5 c7c5 Nodes: 21271672 NPS: 144616 Time: 00:02:27.09 2:27.09 is the time to finish depth 10 on p850 and getting a score at depth 11 takes a long time. Uri
This page took 0.01 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.