Author: Uri Blass
Date: 01:39:24 11/14/01
Go up one level in this thread
On November 14, 2001 at 04:16:57, Gian-Carlo Pascutto wrote: >On November 14, 2001 at 04:08:24, Uri Blass wrote: > >>On November 14, 2001 at 04:00:24, Gian-Carlo Pascutto wrote: >> >>>On November 14, 2001 at 02:32:43, Dann Corbit wrote: >>> >>>>How do I read the distance to mate from this? >>>>There is a mate in 10. >>> >>>My matefinder has no concept of plies, >>>and hence mate depth is irrelevant for it. >>> >>>The solution and first move are guaranteed to be correct though. >>> >>>-- >>>GCP >> >>I guess that it has a tree to prove the mate. >> >>What is the problem to calculate the number of plies in the longest line that >>ends in mate in that tree? > >That value doesn't learn you much. > >If it picks a suboptimal solution, it may find a mate in 100 somewhere >in the tree. > >You can get nearly the same info now by looking at 'MaxDepth', which >is the deepest line it visited. Since it was 27 ply, the mate it found >cannot be deeper than 13 moves. 1)mate in 13 is 25 plies of legal moves. It can be 27 plies from computer point of view only if the computer does not see checkmate by the evaluation function and see the checkmate when the king is captured. If sjeng can see mates only after the king was captured then your 27 plies means mate in at most 13(other programs except chest also do not see always the shortest mate) 2)If you use no hash tables then Maxdepth is an upper bound but the tree may give better upper bound because MaxDepth is also based on lines when mate was not found. 3)If you use hash tables then it is possible that the mate is in more moves than maxdepth suggests because you found a mate in the hash. > >This does not mean there cannot be a shorter mate. It is also the case for the commercial programs and it is clear that when they say mate in 10 they mean mate in at most 10 except cases when there is a bug because of hash tables. Uri
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.