Author: Will Singleton
Date: 19:22:41 03/29/02
Go up one level in this thread
On March 29, 2002 at 22:01:40, Will Singleton wrote: >On March 29, 2002 at 16:39:53, Dann Corbit wrote: > >>On March 29, 2002 at 15:54:08, Art Basham wrote: >> >>>Here white's best move is Bg7..! >>>(with the threat of 2. Qxh7+...and mate by the rook at h8... >>> >>> >>>[D]1r2r1k1/3b1p1p/p4PpB/2qPp3/p1P1R2Q/P7/6PP/5R1K w - - >>> >>>Believe it or not, some engines miss this completely and rather play >>>something like Bc1..etc. ??? >> >>Beowulf took 14.5 seconds on an AMD 950 MHz machine. >> >>Position 1 - White to play >>-------------------------- >>ID = >>FEN = 1r2r1k1/3b1p1p/p4PpB/2qPp3/p1P1R2Q/P7/6PP/5R1K w - - >> >>. r . . r . k . >>. . . b . p . p >>p . . . . P p B >>. . q P p . . . >>p . P . R . . Q >>P . . . . . . . >>. . . . . . P P >>. . . . . R . K >> >>Current Position = 0.58 >>Number of Possible Moves = 36 >> >> 6 98 88 79497 Bc1 a5 Qf2 Rec8 Qxc5 Rxc5 >> 7 118 184 180300 Bc1 h5 Qf2 Rbc8 Bb2 Bf5 Qxc5 Rxc5 >> 8 94 396 420473 Bc1 h5 Qf2 Qxf2 Rxf2 Bf5 Rh4 e4 >> 9 113 1038 1110972 Bc1 h5 Qg5 Rb3 Qh6 Qf8 Qxf8+ Kxf8 c5 Rd3 >> 9 123! 1448 1555074 Bg7 h5 Qg5 Rb3 Qh6 Qb6 Rfe1 Rb1 Rxb1 Qxb1+ >> 9 588 2554 2711402 Bg7 Bg4 Rxg4 h5 Ra1 Qxa3 Qe1 Qc5 Rxa4 hxg4 >> 10 586 4907 5283324 Bg7 h5 Qg5 Qf2 Ree1 Qc5 Qg4 hxg4 d6 >> 11 603 9748 10947729 Bg7 Bg4 Rxg4 h5 Qg5 Qf2 Qc1 Qe2 Rh4 Rb2 g3 Qd3 >>Rf2 >> >>Search Time : 999.0 Seconds >>Total Nodes Searched : 95521440 (84.8% Qui) >>Total Nodes Evaluated : 80995522 >>Cuts : Delta 19222644 : SEE 9400726 : Mate 0 >> : Razor 5045011 : Eval 43111935 >>Extensions : Check 3669449 : OneRep 631520 : CMThreat 119043 >> : Pawn 560266 : Recap 296138 : RevCheck 244444 >>Hash Stores (Size) : 6014980 (2097143) >>Hash Probes (Hits) : 55812193 (4685047) >>EGTB Probes (Hits) : 0 (0) >>Move Ordering : 55.4% >>Best Move = Bg7 >>--Answer: Bg7 ** Correct ** >>Time To Solution = 14.48 sec >>Correct So Far : 1/1 > > >Amateur does well here, probably by luck. Finds Bg7 at ply 4 after 0.12 sec, >score +2, rising to +6 at ply 9, powerbook 300mhz. > >btw, a question re your node numbers. Beowulf reports 95 million total nodes, >but shows only 10 million in ply 11? I assume you just omitted ply 12. > >And, the quiesce percentage seems high, although I notice some programs report >high percentages, and some low. I'm thinking this is due to how nodes are >counted. Does Beowulf happen to count qnodes as calls to quiesce, even if the >quiesce search cuts off prior to making a move? > >Will I just checked Beowulf's source, and indeed it does appear Colin is doing ++qnodes right at the top, before the stand pat cutoff. This is wrong, I think, since it inflates the node count. No real work has been done, no make/unmake. Beowulf is choosing to count a node, even when it decides that the node is not worth looking at (in the case of a check, not even an eval takes place). This results in a wildly different node count than you would get from counting nodes in make. A program that counted nodes in make would show a much lower qnode percentage. Or am I wrong on this? Will
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.