Author: Robert Hyatt
Date: 19:42:00 05/13/01
Go up one level in this thread
On May 13, 2001 at 19:48:59, J. Wesley Cleveland wrote: >On May 12, 2001 at 20:41:23, Robert Hyatt wrote: > >>On May 11, 2001 at 16:50:28, J. Wesley Cleveland wrote: >> >>>Okay. With exact results, you only need the number of plies to the next capture >>>or pawn move stored with each position to solve the 50 move rule problem. >>>Repititions are a non-problem, i.e. if from position A, you know that position B >>>is a forced win, *but* the win leads back through A, you would never choose to >>>move to B, because you would already know there is a shorter win from A. >> >> >>How would you _know_ that either of those positions were forced wins if you >>don't save _everything_ as you search? >> >You know because you have a string of positions in the hash table, each of which >is one ply closer to mate. There *can't* be a repitition, or it would be a >different string. It is just like endgame tablebases, which do not need any >history of positions. I'm not sure I follow. Endgame tables have _all_ positions available during their creation. That is how the algorithm works.. find a position that is marked as "unknown" by backtracking from a position marked as "known". Then you can mark the unknown entry as mate in one more move than the known entry. But you must have _all_ positions stored during the creation... _every_ one.
This page took 0.02 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.