Author: Ed Trice
Date: 05:48:12 01/08/04
Go up one level in this thread
I almost forgot -- I am starting a new checkers project that will probe the perfect play lookup databases in RAM, and UNCOMPRESSED! This program, named ONLY PERFECT, will not have any terms in its evaluations function except material (checker value, king value), the true distance to win, and a means to discriminate difficult draws from trivial draws (the AGGRESSIVE DRAW HEURISTIC.) What I would like to test with this is how well a program will perform postponing losses from the lost 3-move ballots. The lookup will be very fast (nothing to decompress, everything in RAM) and it should prove to be very interesting! Since we learned that there are very long and difficult 4x3 positions, the question is, if the program can elect to play into the longest loss from a great distance away, can it hold other programs to a draw by finding the hardest win for the opponent? If not, will the 9-piece ppl database probed in RAM be required? Or could all programs deal with the 3-move ballot losses? It is an interesting theoretical question that involves "strong solutions" to the game of checkers, which we would be able to filter down the game tree for the first time.
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.