Author: Eelco de Groot
Date: 11:40:12 10/21/99
Go up one level in this thread
On October 21, 1999 at 12:03:37, Georg v. Zimmermann wrote: >1.) this doesn't work cause u can't produce "clones", a different program is a >different program. Yes, but it wouldn't have to be perfect. Even a few percent increase in the number of correctly predicted moves could make a difference... I was thinking more that the program might become a little unstable, for instance if the opponent personality valued positions a little higher on average than the original one, it would be tempted to throw in some sense less moves maybe, to reach an end position were it could apply this evaluation routine with a little higher value. Well that's just what I could come up with. Are there other fundamental objections? > >2.) You can't tell (well maybe if you put enough effort into it but an >"anti-detection" algorithm is too easy to write if it becomes a major problem) >which other program you are playing via autoplayer. I know that Ed Schröder stated that he doesn't want to do this but he said that programs can easily be recognized by the openings they play. Their repertoire is limited because they can't just play any opening successfully, difficult strategies are still out of reach. I know the SSDF already played part of their games where some programs were given information who their opponents were. Regards,Eelco
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.