Author: Robert Henry Durrett
Date: 14:02:03 08/24/98
Go up one level in this thread
On August 24, 1998 at 16:15:45, James Robertson wrote: >On August 24, 1998 at 13:43:45, Robert Henry Durrett wrote: <snip> >While a program's book will have more positions in it related to common opening >positions (i.e. a program will be able to play farther in the book in a Sicilian >rather than a Polish) it is just too much work to find specific tabyia >positions. (1) The tabyia positions are very well known. There is no need at all to search for them. [Although, a semi-automated systematic search would be "interesting," on it's own merits, regardless of how the information were used.] >Why not just feed a computer a pgn file made up of thousands of GM >games and have it convert that into a book? (2) I believe that what you are doing there is creating a tree from thousands of games which have one position in common. That is really what I am suggesting, but for a different position. One could collect all of the chessmaster and GM games having the given [tabyia]position in common, and create a tree originating at that position. This tree would be much smaller than the tree created for the position you are using, i.e. the position that occurs before any move is made in a game. There is no essential difference in method, as far as I can see, between the two cases. >The book theoretically will end up >with the same effect of covering tabyia positions (as you said, they occur more >often in GM games), (3) Actually, I do not believe this is accurate. The "coverage" would be no where near as good. Many tabyias are a dozen moves into the opening and it seems to me that your tree would not be very complete, for that reason. >but the work involved is minimal. (4) Yes, but we get the "nothing ventured, nothing gained" situation. If the programs are to be tested against GMs, they will do better if well prepared for those games. The GMs may very well steer for positions which they understand well, i.e. tabyia positions. >Hope this helps. (5) Thanks. I guess I'm trying to help, too, and maybe to gain a better understanding of chess engines in the process. > >James
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.