Author: Peter McKenzie
Date: 18:13:35 06/26/01
Go up one level in this thread
On June 26, 2001 at 16:23:46, Scott Gasch wrote: Hi Scott, An ideal solution would be to add 2 things to your program: 1) Recognise those blocked positions where it really is hard/impossible to make progress, and adjust the evaluation accordingly. 2) Improve the time management so that it is the human, not the computer that loses on time if such a position is ever reached. I think the first one is the hardest, but a couple of ideas are: - positions where there are 14-16 pawns on the board are more drawish!? - adjust the score slightly towards zero if there are lots of pawns that are either locked (blocked by another pawn), or backward (effectively can't move due to enemy pawn). I don't do any of this in LambChop yet... Peter >Hi, > >I would like to hear about philosophies about anti-human code and ideas for how >to implement it. I've decided I need to do something after watching my engine >lose on time in an utterly blocked position in a standard game... to an opponent >who is moving his king back and forth or shuffling pieces around. But I do not >want to simply add code to eval that says "blocked positions are terrible" for >fear of decreasing engine strength. > >One idea is to realize you're in a blocked position and spend very little time >thinking... so that the engine doesn't get in time trouble. Think longer only >if the PV move is a capture or pawn push. Another idea is obviously to >recognize blocked positions in eval and penalize for them... What are good >criteria for identifying a blocked position? > >I wonder if there any way to understand the critical squares of a blocked >position or to understand a blocked position better... > >Thanks, >Scott
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.