Author: Gian-Carlo Pascutto
Date: 09:52:06 06/20/00
Go up one level in this thread
On June 20, 2000 at 09:10:40, Robert Hyatt wrote: >Ignore this. To parse a move, I do the following: > >generate all moves > >use the input text to eliminate moves from the list, one by one. IE if the >input move starts with N, then I eliminate everything but knight moves. I >then parse source/destination/promotion/check/etc and eliminate moves that >don't fit the requirement. If I end up with one move, it is _the_ move. If >I end up with zero, the move parsed is illegal. If I end up with more than >one move left, the move parsed is ambiguous. My explanation must have been insufficent, as this is the same as Faile/Extract do. The only difference is that they first parse in the source/dest/etc and then do the elimination, not after every step. The difference in efficiency is very minor I think. >In Cray Blitz, we built the book by parsing each move, and then doing a short >search from each resulting position to be sure we weren't walking into obvious >blunders. If a score was exceptionally low or high, the search was extended >to a couple of minutes if needed. It took us a few days to build our book >because of this, but we didn't do it very often. And yes, that was _days_ as >in 24 hour non-stop computing. But was it worth it? A related question: could you explain how Crafty's binary book building works? I've tried figuring it out from the source, but it was a bit over my head. -- GCP
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.