Author: Robert Hyatt
Date: 18:46:07 09/14/99
Go up one level in this thread
On September 14, 1999 at 19:12:17, Bas Hamstra wrote: >Bob, > >We had a discussion about best moves. It became clear that you can not choose a >best move from scores < Alpha. I store always a move in the hashrecord, as extra >check if the hashrecord is legal. So when an UPPERBOUND record comes up, I throw >away the hashmove, because it is no real best move. > >But isn't it an idea to keep a seperate flag BM in the hashrecord? You could >keep track of a "bestmove" flag apart from the type (boundtype) of the >hashrecord, I think. Just don't overwrite the bestmove in a hashed position if >there is no new bestmove, only adjust the bounds. Keep the old bm. Idea? Old >idea? Nonsense again? :-) > >Another peculiar thing: if there is no real best move (al fail low on alpha, >including the first) but I store the first move and pretend it is a bestmove, I >get better results than not storing anything at all. IE considerably less nodes. > > > >Regards, >Bas Hamstra. No... this is an OK idea, and is one I suggested in the other thread about doing this... ie if you find a duplicate signature, but the best move is present, save that best move in this "upper" hash entry (since none is available). After all, that 'best move' had to come from somewhere, and it probably makes sense...
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.