Author: Gian-Carlo Pascutto
Date: 02:24:27 10/06/01
Go up one level in this thread
On October 05, 2001 at 12:40:31, David Rientjes wrote: >I often tell people on chess.net that Compablanca does not have "books" like >other computers do. Compablanca has a set number of openings which it follows >but usually after move 10 (or whenever an opponent deverts from an opening), >it's on its own. And this would be different _how_ ? > Compablanca uses a list of priorities which it attempts to >maintain which start as follows: > > 1. prevent a checkmate or a positional advantage in which a checkmate may >become possible; > 2. exploit a checkmate or a positional advantage on the opponent and disregard >material advantage possibilites; > >Following that, there are a list of priorities for positional advantages, >material advantages, promoting of pawns, king usage, etc. The list of >priorities is obviously not discussed by myself since that would open >Compablanca up to be exploited if someone were to know this list. Actually, this is also how everybody elses engine works. (evaluation parameters with different weights) >There have >been several human users who have had material advantage of three or four pawns >against it and have even gone on to checkmate it (a national master comes to >mind who checkmated it in less than 40 moves). Compablanca is often vulnerable >when playing someone who is capable of making very precise and important moves >in comparison to other players who can make one or two slight move errors which >open Compablanca's FOA (field of attack) and allows him to eventually go on to >win. Again, nothing new or exceptional here. >Compablanca has not only been beaten in standard matches, but also in wild >matches, and pawns only matches (where a computer would normally dominate). Hmm, computers usually have great trouble in pawn endgines, unless you do all the stuff crafty does. >Compablanca is not open to matches from other computers since it would obviously >suffer great vulnerability. Compablanca was created within the past five or six >weeks (the first week being the most work-filled) by a single person. Other >engines like Crafty, Fritz, etc. (engines I have very little experience with >myself) have been worked on for years by a team of people. The disadvantage to >Compablanca is evident. aka I don't play others because they are stronger? Weird attitude. >My hope for Compablanca was that it would provide the users of chess.net with a >powerful engine in which they may play against to improve their chess games. >Rating restrictions are a new thing (about a week old) to prevent inevitable >problems in which Compablanca would forfeit for lagging too much or >disconnecting. Up to last week, all users were open to play against it. So basically you do care about its rating. >Again, my goal is not to promote myself as a great programmer or a great chess >player by any means. My goal is to provide good chess players with a tool to >possibly improve their chess games. I don't care to spend my time promoting it >or bragging about its successes. Neither do we. >Remember, I will not be back to this message board to view the responses to this >message or check-up on future debates about Compablanca (which I whole-heartedly >welcome) or any other chess engine. I was only invited here once by the >original author and was told to "defend myself". I don't believe I need to >defend myself in any way since very few people know I wrote the engine and >everyone else seems to enjoy playing against it more than debating whether a >single person actually wrote it or not. Oh, I don't think the debate is about whether a single person wrote it or not, but rather if you wrote it at all, from scratch, in 3 to 4 weeks. Based on your post, I have my doubts. BTW. I _do_ think its possible to make a 2600 engine in a few weeks, and can think of at least 2 ways: a) You have already written a much stronger engine once and start over b) You read up on how bitboards work and copy all the stuff you need out of crafty -- GCP
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.