Author: Frank Schneider
Date: 23:45:34 08/22/03
Go up one level in this thread
On August 22, 2003 at 10:54:17, Steven Edwards wrote: >The following position is #398 from _Winning Chess Sacrifices and Combinations_: > >[D]2qrr1n1/3b1kp1/2pBpn1p/1p2PP2/p2P4/1BP5/P3Q1PP/4RRK1 w - - 0 1 > >It has been referenced in a number of CC papers over the past three decades >including the description of the NWU program Chess 4.x by Slate and Atkin. It >was also a test position for the MacLisp Paradise program by David Wilkins at >SRI. > >The position is a mate in ten. My old program Spector, moribund for seven >years, picks the key move 1. Qh5+ at iteration eight as "obviously winning" and >sees the full mate PV on iteration nine. > >PV: Qh5+ Nxh5 fxe6+ Kg6 Bc2+ Kg5 Rf5+ Kg6 Rf6+ Kg5 Rg6+ Kh4 Re4+ Nf4 Rxf4+ Kh5 >Rg3 Re7 Bg6# > >Spector requires over fifty million nodes for the complete search and I'll guess >that the numbers aren't all that much different for most programs using common >search techniques. > >BUT, the above mentioned Paradise chess program pulls out the mate PV with a >search tree containing only 109 nodes. It solves many other tactical puzzles >with similarly small search trees. > >Are there any programs out there today that emphasize knowledge per node rather >than nodes per second? Gromit 3.10.16 (predecessor of Anaconda) on an iPAQ 3660, 8MB Hash needed 35sec to find Qh5 and announces a Mate in 10 after 68sec (iteration 7), which is about 320000 nodes. Frank
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.