Author: Stuart Cracraft
Date: 15:15:48 03/01/99
Go up one level in this thread
On March 01, 1999 at 16:35:27, Dan Homan wrote: >On March 01, 1999 at 16:15:05, Stuart Cracraft wrote: > >>On March 01, 1999 at 15:25:45, Will Singleton wrote: >> >>> >>>On March 01, 1999 at 13:28:50, Stuart Cracraft wrote: >>> >>>>Hi, >>>> >>>>I'm curious what kinds of things people have for evaluating king safety. >>>> >>>>Yes I've looked at Crafty but am more interested now in a variety of >>>>people's viewpoints on KING SAFETY. >>>> >>>>Stuart >>> >>>Mine isn't anything to shout about, but it's a lot better than it used to be. I >>>used to fall for just about any kside attack a good player could come up with; >>>now that occurs more rarely. >>> >>>I talked with Dan Homan about this some time ago, and he shared some ideas about >>>his methods. I've modified the approach since then to fit the way my program >>>works. Basically, I just compute two things: how much is the king being >>>threatened, and how much do I care about it. That is, if my king is >>>well-protected, should I care if the opponent has a couple pieces nearby? >>>Certainly not as much as when my king has a pawn missing and no defensive pieces >>>around. >>> >>>So I just multiply the threat by the "do I care" factor, and that seems to work >>>pretty well. It was difficult to balance the numbers (attacker vs defender), >>>resulting in my program throwing pieces at the enemy sometimes. But I do it >>>asymetrically now, and most times it works. >>> >>>Will >> >>And how do you compute "how much the king is being threatened" and > >How many pieces from the other side are nearby? Which pieces and >how close? Do you have an open files next to the king? Are there >Rooks on any such open files? > >>"how much you care" ? > >Is the king protected by a complete/partial/no pawn shield? Are >other of your pieces nearby to help in case there is trouble? And >anything else you can think of.... > >This stuff is all pretty crude, but it is easy to calculate. >(The exact details will, of course, depend on the particulars >of your program.) The main thing is that it seems to work most >of the time. > >My program uses a combination of static scores for obvious >vulnerabilities that I would like to avoid in general and >some of these >"how much do I care"*"how much the king is being threatened" >terms that can alert me to more dynamic dangers. > > - Dan Okay, then let's get more into this "how much do I care" factor. I presume you care more when the enemy queen is around or it is not the endgame. Any other times you don't care?
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.