Author: Robert Hyatt
Date: 05:46:59 01/21/99
Go up one level in this thread
On January 20, 1999 at 19:24:50, William Bryant wrote: > >> >>IE I have a variable (array) hash_move[ply]. When I call HashProbe() if it >>gets a signature match it copies the best move to hash_move[ply] while it is >>checking the draft and so forth. If the lookup is useless, search calls >>NextMove() to choose the first move, and the one I try is hash_move[ply] >>(making sure it is still a valid move here). And I get to do this _before_ >>I generate any moves at all... > >What advantage is there to making the hash move 'before' preparing the >move list for this position. If it fails high, then I realise that a move list >is not needed, but if it fails low or returns a new alpha score, you will still >need to search the rest of the legal move of this ply, right? > yes.. but what if your 'hash move' causes a fail high 25% of the time. Wouldn't you like to speed up your move generator by 25%? You can, simply by not calling it. :) > >Also a few last quick questions. > >If a current position has no legal moves (checkmate or stalemate), is the >mate value stored in the hash table as an EXACT score? MATE should always be exact. But you have to 'adjust' it... to be mate-in-N from the current ply, not mate-in-N from the root. In your case, MATE is fine since since that is absolute here... > >When a position leads to a draw by 50 move rule, the return value will be 0, but >since arriving at this position by a different route may not create a 50 move >draw, do you store the positional evaluation score in the hash table even though >you are returning a value of zero? No. I store the 'score' which is zero in this case. Yes it is wrong. But if you ignore storing draws, it is _still_ wrong, because hashing has no 'path' information in the signature. So just because you store +3.0, doesn't mean there isn't a path that reaches that position _after_ a 3-fold repetition. So errors are going to occur no matter what happens... > >As always, I greatly appreciate you assistance. > >William >wbryant@ix.netcom.com
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.