Author: Robert Hyatt
Date: 19:27:26 02/16/99
Go up one level in this thread
On February 16, 1999 at 22:04:23, William Bryant wrote: >Dr. Hyatt, > > Following you discussion of checks in the Qsearch in rgcc, I reviewed >Quiesce() in the crafty source. You calculate a delta value as follows: > the delta = alpha - 8/10 of a pawn - the difference in material > >Do I understand correctly that, if the value of captured piece is not at least >as great as this delta value, the there is no point is searching this capture >and no point in adding it to the qsearch move list? yes... the assumption is that a single capture can, at best, improve the score by the amount of the piece being captures, plus maybe an additional pawn in compensation from the Evaluate() function (note, positional scores can be much bigger than a pawn, but this case means that the _capture_ has to result in a positional swing of _another_ 1 pawn which is very unlikely; because the alpha value in q-search is already set to the _current_ positional+material score... If the capture still leaves the score way below alpha, there is no use in following it further. > >Also, do you sort the captures by the least valuable attacker / most valuable >piece with the idea that pawn capturing a queen will most likely generate a >cutoff meaning that you can skip the rook captures knight move? No... If the captured piece is more valuable than the capturing piece, I put these first, otherwise, I use my static exchange evaluator to carefully play out the captures on that square to see how it ends up... and I use that result to sort the moves. > >Thank you. > >William Bryant >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.