Author: Robert Hyatt
Date: 07:46:14 06/09/03
Go up one level in this thread
On June 09, 2003 at 09:54:24, Mike Adams wrote: >once again going over my hashing code making some changes and fixing bugs as >well as trying to make sense of it all. Just a couple of questions. i store the >hash move, from to and depth searched, i store fail high, fail low or value if >it changed alpha. Of course i store the original hash number as well. Is that >enough right there? And yea to save space i store fail high low and value in one >variable which i can parse to figure out what it is. > >basicly if it meets the depth test and is the right actual hash number I return >a score beta for fail high, alpha for fail low or i return the value if it wasnt >a fail at all and if the value is between alpha and beta otherwise if its not a >fail but he value is to high or low i return alpha or beta. > >People who do hasing what do you think. this sound correct. > >thanks >Mike Adams My only question concerns the bounds. If you search and you get a fail high, that beta value gets stored in the table as a _lower_ bound, because the score is at _least_ that high and possibly higher. When you get a hit later, and the depth remaining (draft) meets your requirement, you need to compare the current lower bound from the table to your current beta value. If your current beta value is _higher_ than the table lower bound, you can't take a "cutoff" path. If your current beta value is <= the table lower bound, you can. The inverse holds true for the fail low case where you store an upper bound in the table..
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.