Author: rasjid chan
Date: 10:22:48 01/17/04
Go up one level in this thread
On January 17, 2004 at 12:33:53, Robert Hyatt wrote: >On January 17, 2004 at 02:38:24, rasjid chan wrote: > >>On January 16, 2004 at 16:31:17, Robert Hyatt wrote: >> >>>On January 16, 2004 at 12:07:00, Tord Romstad wrote: >>> >>>>On January 16, 2004 at 10:02:39, Robert Hyatt wrote: >>>> >>>>>I would worry about overflow there. >>>>> >>>>>IE e4 could be attacked by pawns on d3 and f3, a bishop at d5 with a queen >>>>>behind it (4 so far), plus two rooks and two knights. That's 8. It just >>>>>overflowed to zero. Not to mention the promotion problems where there is a >>>>>third knight or bishop or second queen... >>>> >>>>It is not so dangerous if you are just a bit careful about where and how >>>>you use the information. I use it mostly in the evaluation function, in >>>>order to compute stuff like mobility, space and king safety. That one >>>>or two squares on the board are evaluated incorrectly once every hundred >>>>nodes or so usually does not have any catastrophic consequences. >>>> >>>>You are of course right that an SEE based on this simple kind of attack >>>>tables is not reliable enough to be used to replace a qsearch. For this >>>>purpose, I use a much slower and more accurate SEE. >>>> >>>>Tord >>> >>>I was more worried about the actual overflow. IE if this is a char, 8 attackers >>>appears to be zero. If it is packed into a word, then 8 attackers wraps into >>>the next char to the left and corrupts that... >> >>I think there is no overflow. His BIT0 is the LSB. 8 attacker overflows to >>pawn attacking so almost it will be true. having 2 queen almost mean game >>decided. Extra Knights or Bishop has other simple solution. > > >The game might not be so decided if your extra queen screws up evaluation and >capture analysis. :) Of course I cannot be too serious. My program is far far away.
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.