Author: Robert Hyatt
Date: 13:43:41 07/23/99
Go up one level in this thread
On July 23, 1999 at 15:40:52, GBes wrote: >On July 23, 1999 at 11:42:07, Scott Gasch wrote: > >>Hi. >> >>I have implemented hashing (transposition) tables based on crafty's code and >>some posts about it here. It makes sense and works very well... but I am having >>trouble with one aspect of it. >> >>In a position where the side on move has a very good move (promote a pawn, >>capture a queen, etc) I am hitting the hash table at every depth of my search >>and the table is returning a bound instead of an exact score. This happens, for >>example, on plys 1 through 6. The program then begins to search ply 7 but runs >>out of time. Since plys 1-6 were bounds from the hash, I have no move at this >>point. > >Don't know if i understand well your problem...(still struggling with hashtables >too) >Seems to me that every hit to the hashtable should you give a SCORE ( exact >lower upper worthless ) plus *always* an associated move, even if the score is >just a bound. this works for me. > >good luck, >gerrit bes actually you only can get a 'best' move from a "LOWER" bound (fail high) position. In an "UPPER" bound (fail low) all moves were bad and there is no 'best' one. But the problem seems to be terminating searching when he gets a "LOWER" hit... that tells him the score is > current beta, but you do _not_ stop there. He should run beta to +infinity and then search again. Now that bound will be worthless since it is < beta and the search has to find the real path to the best position...
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.