Author: Gordon Rattray
Date: 07:35:29 07/20/01
Go up one level in this thread
On July 19, 2001 at 16:49:11, Otello Gnaramori wrote: >Excerpt from the Chessbase page of the latest report from Dortmund : > >http://www.chessbase.com/events/dortm01/dortm03.htm > ><snip> >"In the Man vs Machine match Dr Robert Hübner, playing with the white pieces, >offered his opponent Deep Fritz a draw after 21 moves. The operator accepted, >even though Fritz thought it was slightly ahead in the game. The score between >the two is now 2:2. The last two games will be played on Saturday and Sunday." ><snip> > > >Probably when the operator pressed the button "Offer Draw" Fritz answered "I >think that the game is still quite interesting , let's play on", but the human >operator didn't want to disappoint Dr. Hubner. > >Regards. Responses so far have included two lines of thought... Firstly, the practical point of view that a human operator sometimes needs to save any unnecessary "boredom" by helping to terminate the game much earlier than the computer would agree to. Alternatively, there is the reasoning that it is the computer that is playing and not the operator. Hence, the operator should have as little influence on the game as possible. My opinion is that while the current situation may require the operator to get involved for practical reasons, programmers should aim for their programs to be independent as possible. Just like human players, programs should become known for their judgement of when to resign or agree to a draw. This characteristic shouldn't be covered up by letting a human make the decision. If "program X" becomes disliked for its reluctance to terminate play when such a decision is the right one, then so be it. This may affect its attractiveness for participation in human competitions, but if so, the issue should be fixed, not worked around. For programmers, it will be necessary to avoid the other extreme: resigning or agreeing to a draw too early. But computer chess is full of such problems, and working around them is not a solution. The operator should be responsible for setting up the program prior to the first move. Thereafter the computer should make all decisions until the game is over. Similar to a "draw contempt" factor, the program should support the ability to be configured for a particular strength of opponent. The benefit is that in cases where the computer refuses a draw, when the operator would have accepted, but then goes on to win or lose, the result is a true reflection of the program. If programs aren't clever enough at the moment, I won't disagree to practical human interventions. But looking to the future, I'd rather the programs were improved in this area. Gordon
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.