Author: KarinsDad
Date: 14:56:34 03/31/00
Go up one level in this thread
On March 31, 2000 at 12:16:28, Colin Frayn wrote: >>You might be surprised at the knowledge even in tiny chess engines. Look at >the very clever eval code in Phalanx, for instance. > >Substitute 'incomprehensible' :) >I'm sure it's very clever, but it would take hours of work to find out what he's >doing. > >I tried to put quite a bit of chess knowledge into ColChess, but it's not >exactly the greatest analysis function ever written. Hopefully it's >comprehensible though as I used 8*8 arrays to store the pieces rather than >bitboards or any of that nonsense :)) >(I hope Bob Hyatt isn't reading this) Are you implying that bitboards are incomprehensible, or just incomprehensible to you? > >I think that improving the static eval function would improve the performance of >my program a lot more than slghtly tweaking its tactical speed. I could >convincingly argue with anyone who says otherwise. Having said that... it's >important to not lag _too_ far behind in sheer search depth simply because >you'll get outplayed with tactics every time. I have come to the conclusion that the best move is the best move is the best move, regardless of how it is found. In other words, if a program is brilliant at positional play, but only searches 8 ply deep, it could still conceivably beat a program that searches 12 ply deep since it often makes moves that GMs would make. I have this theory that good understanding of the principals of chess would enable a program to automatically avoid the pitfalls of given positions due to the avoidance of those positions in the first place (e.g. if you program does not play a gambit, it never has to pick the pawn back up). KarinsDad :)
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.