Author: Uri Blass
Date: 10:36:08 06/24/02
Go up one level in this thread
On June 24, 2002 at 13:10:49, Gian-Carlo Pascutto wrote: >On June 24, 2002 at 11:54:55, Steve Coladonato wrote: > >>The nominal depth may be "artificial" because of the extensions and pruning but >>would it not be a gauge of how well the alogrithms/heuristics are doing in the >>evaluation? I understand from the other replies that Junior counts "plies" >>differently, but do not the other programs use the term ply equivalently (I >>don't know if that's a word or not)? > >No. The amount of work that is done for each ply is different for each >program. Plies are not comparable. Plies are not comparable but this was not the question of the poster(he said that he understands that the nominal depth may be artifical. Ply has a clear definition and programs(except Junior has the same meaning of ply). If a program searches x plies forward then it searches a lot of lines x plies forward(it can search lines more than x plies forward if there are extensions and it can search lines less than x plies forward). > >>I do know that if you chase an evaluation >>down the variation the score can become much different near the end of the >>variation. So would not a deeper ply depth also be an indicator for the >>correctness of at least the initial part of the variation? > >You cannot compare plies, hence your last question makes no >sense. > >-- >GCP You can compare plies of the same program and if a program searches deeper then in most cases you can expect a bigger part of it's pv to be correct. The only exception that I know is programs who use some algorithm that I do not understand today and hash tables to construct their pv but i believe that most programs do not do it. I understood that Fritz7 does it and not Deep Fritz. Note that I do not like not knowing the pv. Even tscp use the pv for better order of moves and I can imagine that it is possible to use the pv for other purposes(for example to decide to extend pv moves in part of the cases) so I doubt if it is good to use some method that does not let the program even to know the principal variation. Uri
This page took 0.01 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.