Author: Robert Hyatt
Date: 16:50:29 11/28/02
Go up one level in this thread
On November 28, 2002 at 11:13:22, Mike Adams wrote: >Just a few quick questions to make sure i have the basics of hashing down. When >hashing fail high, >=beta and fail low <alpa you only use the info to not search >next time if hash depth>=current search depth? and this is the same rule if you >hash scores say alpha changes? return hash score wich is your previously >searched alpha if depth on hash entry >=current depth searched?. And play the >hashed move first if it had changed alpha but the depth is not enough to use the >score? And finnaly if hashing in qsearch use -depth and any qsearch hashed >postion always play a hashed postion in qsearch if the node had also been >searched in both search and qsearch before becaue it automaticly has more depht >and is reliable? thanks, just tyring to make sure i'm covering the basics. Mike The idea is this. If you search at a ply and get a bound (score >= beta) then you store a "LOWER" flag and either the score or beta depending on how you are doing this. (this is a lower bound, as you know the score is >= to this value). If you get a hit later, all you have to be sure of is that this "bound" is useful. IE you take the value you stored in the table, and if it is >= the current beta bound, then the cutoff is valid and you don't search further. If it is < currrent beta, you can't take the cutoff and you simply search normally. You do the same for the case where you search all moves and store "alpha" as an "UPPER" bund since you know the score is <= alpha, then alpha is the largest the score could be and it is probably even lower. If you get a hash hit and the depth is ok, you have to make sure that the table bound is <= the current alpha value and if so, you can take the cutoff. EXACT should be pretty obvious...
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.