Author: martin fierz
Date: 20:40:42 03/14/02
Go up one level in this thread
On March 14, 2002 at 23:23:09, Tina Long wrote: >Hi Guys, >Bobby Fischer invented a version of checkers called Fischer-Random-Checkers. >The four pieces on each back row are shuffled randomly (although piece two must >remain on the left of piece three). >This is to eliminate the dependance of human and machine players on opening >theory. doesnt seem to make a difference :-) >My questions to Martin: >At the start, how deep (how many plies) do the best Checkers programs, on >average, play by "Opening Book" before they start to "think". there is no definite answer to this. my computer generated opening book has variations in which go up to 60 ply deep, and also has variations in which end up in positions which it can prove to be a win/loss/draw by endgame database >In an average middlegame move of 2mins (?), how deep do the best Checkers >programs look - I would guess 30 or even 40 ply. it depends heavily on the position. something like 27 ply nominal search depth is normal. with extensions up to 45 ply. >Do the best Checkers programs use "Tablebases" (when there are few pieces left, >hardcoding that states "in this board position make this move") instead of >"thought". yes. the standard was with 6-pieces for a long time, because the chinook team published this database for free use. they had an unpublished 8-piece database. now, two commercial programs have the 8-piece database too, and one team checked their db against the chinook db and they both found mistakes :-) schaeffer of the chinook team is computing the 9- and 10-piece db and i think he wants to commercialize them. handling these dbs is rather tough though, the 8-piece db compresses to about 6GB, and the 10-piece db will need much more space. aloha martin
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.