Author: Harald Faber
Date: 02:35:21 03/20/98
Go up one level in this thread
On March 18, 1998 at 15:58:40, Bruce Moreland wrote: >>It can only work IF SSDF starts testing COMMERCIAL AVAILABLE PROGAMS, >>meaning not only available but also bought! >>So there can't be possible cheats or hidden autoplayers etc. >>Only test what anybody can buy at the store or whereever. >> >>THAT is fair. > >You can never have a "fair" list as long as the new version of program A >(with learning or with a lot of human opening book intervention) has >played 5,000 autoplayer games against the current version of program B >(with or without learning) in the lab, and then programs A and B are >played against each other by the SSDF people. Doesn't it depend on s.th. else? I mean if the learning is implemented in the program you are right, if not (for it is saved in a learning file like MCP does) there shouldn't be any advantage because bought programs have no learning files coming with, or if they do, the testers should delete em or disable (by renaming etc.) >The one who releases most recently has an advantage. >Intensive training against specific opponents screws everything up. >bruce Wasn't it so that advantages against 1 program could give disadvantages against others?
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.