Author: Dan Homan
Date: 16:46:47 01/15/99
Go up one level in this thread
On January 15, 1999 at 16:10:53, James Robertson wrote: >On January 15, 1999 at 13:43:40, Dann Corbit wrote: > >>GNUChess is a *pansy*. >>Try it against a real program. >>A 3 or better score in a ten game match against commercial programs or crafty >>would be impressive. > >I resent that. My program has a score of 0.0 in about 20 games with Gnuchess. > >James Hi James, I remember my program losing to a RadioShack 1650L repeatedly early on. The thing had the cpu power of a toaster and it was 6 months before my program running on a 486 could beat it on level 1. Gnuchess can certainly seems like a monster at first. I started my program in the fall of 1996 and it was a full year before EXchess could even play a reasonable looking game with GNUchess. No wins or draws at that point, but at least it played what looked like a reasonable game occasionally. Soon after that (and a re-write to undo some of the handicapping mistakes I had made early on) my program became increasingly more competitive with GNU. The most recent version scores slightly better than 50% against it in my informal testing. The biggest thing to remember in competiting with GNU chess is that it has pretty solid code (meaning few if any bugs) with a pretty well tuned eval. When I was first competing with it, I thought "there must be some big search trick or eval trick that I am missing.... Why can't I compete?" The major reason was nothing flashy... rather it was bugs that were beating me. Bugs in my search and bugs in my eval. I went on a lengthy campaign of bug detection/squishing. By far, squishing bugs provides the largest ELO return on time spent. - Dan
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.