Author: Robert Hyatt
Date: 10:42:49 02/20/02
Go up one level in this thread
On February 20, 2002 at 12:34:30, Gian-Carlo Pascutto wrote: >On February 20, 2002 at 12:10:04, Robert Hyatt wrote: > >>Whatever it takes to survive. Better to defend and win than to attack >>and lose... which is _very_ common against a few particular players on >>ICC. >> >>Why open up both yourself _and_ your opponent and hope _you_ get the attack >>going first? Humans are _better_ at analyzing such positions than any computer >>program around. Playing right into their strength is incredibly dangerous... > >I consider it a challenge to make some nasty surpises for them >when they try to play towards their strength :) So instead of >avoiding stonewalls, I've added code to evaluate them. It doesn't >always work of course, but it does generate stuff I'm proud of >regularly. Especially if the opponent is one of those 'stupid' >other programs that doesn't understand stonewall-like positions :) > >About the original point: I'm referring to that code specifically >because it will really _hurt_ you in certain positions. Sometimes >the right plan _is_ to attack, and it's not so uncommon to see >Crafty just sit and get crushed at those times. > >I'm wondering if you've ever experimented with adding some code >to automatically reduce that assymetry when playing a computer >opponent or an opposite-sides castled position. > >-- >GCP Many times. Crafty has a good idea of when to attack and when to not attack. You have to look at (and understand) the second-order and third-order stuff done in the king safety code. IE if it can get pieces to the king-side first, then it will certainly open things up. If it can't, or if it is not sure, it won't... I specifically try to avoid castling opposite for that very reason as well... Typically a single tempi is all that decides the game. and it is just easier to avoid the risk, based on a lot of testing. YMMV of course, but my king safety weights have been developed over hundreds of thousands of games... And I have tweaked them up and down from time to time to see what effect it has...
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.