Author: Mark Young
Date: 14:09:12 09/08/98
Go up one level in this thread
On September 08, 1998 at 16:07:31, David Eppstein wrote: >While I don't disagree with your judgement in this game, I have some concern >with your methodology: > >On September 08, 1998 at 14:02:44, Dirk Frickenschmidt wrote: >>In the game below Rebel9 played it's Slav >>defence very well, finally reaching an endgame with one pawn up. I already was >>convinced that Junior would suffer its first loss against Rebel. >... >>I judged the game as win for Junior >>after move 65; if anyone disagrees it will probably be easy to demonstrate the >>win: black cannot prevent to lose one remaining pawn after the other and then >>ist mated with king+rook against king. > >What is the difference between these two situations? >In one, you are convinced that Junior will lose, so you keep playing. >In the other, you are convinced that Junior will win, so you end the game. > >In what circumstances do you declare a game over? >Shouldn't you wait until the programs themselves checkmate or resign? >Otherwise you leave yourself open to questions of partiality. It’s very hard not to pull for one program or another when testing. There are many ways to consciously and unconsciously manipulate results when testing programs. This may explain why people tend to get the results they want when playing computer Vs computer games, and only one person is running the tournament and playing the games. As imperfect as SSDF is, it is best to have the programs play each other with as little human intervention as possible. And play many games against each program not just a few. Still any testing is fun and interesting to follow.
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.