Author: Russell Reagan
Date: 13:40:07 04/04/03
Go up one level in this thread
On April 04, 2003 at 16:11:48, Dieter Buerssner wrote: >Only for the FEN Standard, the ep square must be e3 here. A very doubious rule >(I think it is wrong) in that Standard. For the search of an engine (for example >for hashing), there is no need to set/hash the ep square, when the pawn cannot >be captured. And then your problem is gone ... One might still argue, that in >the few real ep-cases, hash could even help when you ignore ep. Perhaps - I >don't think, it don't think you could gain much. > >BTW. Setting the ep target and hashing it always yields in bugs in one typical >method of repetition detection (comparing with previous hash signatures): When >the position is on the board the second time, you won't see it ... Not hashing >ep target could yield in similar bugs. Setting ep target only when a ep capture >is possible will solve those problems. However, one subtle problem could remain: >it seems that the pawn can be captured ep, but it really cant't be captured >because the capturing pawn is pinned to the king. This might be rather expensive >to detect in the inner loops of the search. Very good points Dieter. I wonder if a different kind of hashing would work here, involving not a hash of the position, but of the moves. A key could be computed during each movegen involving the from and to squares, along with the piece type and maybe a move type. Are there positions which are different, but have the same moves? I can't think of any. I wonder if it would matter if you hashed pseudo-legal moves, or if they would have to be 100% legal. I guess this method of hashing would also be a bit slower, assuming it would work in the first place. Any thoughts?
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.