Author: Christophe Theron
Date: 21:31:21 04/27/00
Go up one level in this thread
On April 27, 2000 at 14:50:40, Robert Hyatt wrote:
>On April 27, 2000 at 08:50:02, Mogens Larsen wrote:
>
>>On April 27, 2000 at 07:42:12, Robert Hyatt wrote:
>>
>>
>>>This discussion has happened before. The idea has its plusses and minuses.
>>>
>>>The main plus is that it eliminates 'cooking a book'.
>>>
>>>The main minus is that even if a program would _never_ play a particular
>>>opening in real games, because (perhaps) it isn't well-tuned for some types
>>>of pawn structures, it still has to play the position in this particular
>>>game. Humans don't play like that.. IE I won't play the french as black
>>>because I don't like the cramped position. I won't play the Colle-type
>>>positions for the same reason. Ditto closed sicilian as white.
>>
>>My point exactly. I've seen enough computer chess programs to believe that the
>>chess knowledge of a program should be reflected in its opening book, ie. open
>>vs. cramped positions.
>>
>>>It is an interesting test, maybe, although I have never looked at the positions
>>>to see what they look like, but it really doesn't say much about how real games
>>>will be played, since _everybody_ gets to pick their own openings in real games.
>>
>>Neither have I, but it's entirely unrelated to OTB chess IMHO.
>>
>>Best wishes...
>>Mogens
>
>
>It isn't horrible.. It is just the opposite extreme. On one end, you have
>cooked books and killer books. On the other end you have pre-determined start
>positions. OTB is somewhere in the middle...
I suggest that instead of using Nunn positions we take starting positions that
are in the opening books of BOTH programs.
This would eliminate book randomness, book cooking, avoid the positions that a
given program would not play, and offer reproducibility of the experiment.
This is not aimed specifically at the Fritz-Crafty case. I think it's an
interesting experiment for any program pair.
Christophe
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.