Author: Ricardo Gibert
Date: 04:56:19 02/03/05
Go up one level in this thread
On February 02, 2005 at 13:25:20, Dieter Buerssner wrote: >If anybody with a rather big machine wants to solve this fast, one idea. Use two >hash tables - one for the pawn endgame positions and one normal. Actually, the >first one would not be a hash table at all, but more like a TB. About 2 GB of >memory would be enough. As long as we have a pawn endgame (or when we have one >again after perhaps a small intermezzo with a Q endgame) we can index all >possible positions. We have 6 files with pawns. On each file, there can only be >one pawn or none. So each file has 7 possible states (pawn on 2nd row, 3rd, ... >7th, no pawn). There are 3612 legal KK positions, so all pawn endgames need > >3612 * 7^^6 = 4.2xe8 These can't be "all pawn endgames" since a pawn capture of a piece from a pawn promotion is possible. > >A move can be stored in 1 byte (for these positions) easily. Score->2 bytes. 1 >byte depth. This is less than 2 GB then. Remaining RAM can be used for the >normal HT. One would only store white to move positions. For black to move, >rotate the board by 180 degrees and switch white to black. > >Equipped with this, I'd try a search with infinite extensions while the position >is a pawn endgame. Perhaps 2 bounds are needed - then 2 GB would not be enough, >and some expensive hardware would be needed (I think). > >Regards, >Dieter
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.