Author: Ricardo Gibert
Date: 02:30:45 07/27/02
Go up one level in this thread
On July 27, 2002 at 05:01:42, Jesper Antonsson wrote: >I thought of a simple idea of EGTB optimization and I wonder whether it's been >tried already. Imagine we have a position that is legal on the surface, but >unreachable in a real game, like this one: > >[D]8/8/8/8/8/4k3/4p3/4K3 b - - 0 1 This position is reachable in a real game. Whites previous move was a move to get out of check by the pawn. Blacks previous move before that was a capture by his pawn giving check, etc. > >I believe standard EGTB calculation algorithms will store this as a mate in n >for black, as Kf3 or Kd3 result in wins. Wouldn't it be beneficial to mark such >positions as broken, so as to get better compression? It can obviously be done, >even for "deeper" unreachable positions, essentially by iterating over the EGTB, >but perhaps it requires information about illegal positions that aren't in >current EGTB, and then a regeneration would be necessary. > >Note that for five men tables, just the example position above will come up in >many variants with two more black pieces on the board. I have no idea of the >ratio of such broken positions to legal positions, though, but it would be >interesting to know.
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.