Author: Robert Hyatt
Date: 20:44:17 10/12/04
Go up one level in this thread
On October 12, 2004 at 21:17:34, Duncan Roberts wrote: >you once said I think that crafty in a quad opteron, searches anything between >16 -21 ply. > >Would there be any point in crafty having a function which would give an average >ply search for the whole game. At least this way you would have a way to compare >different versions of crafty, and same versions of crafty on different hardware, >regarding ply search. I'm not sure it would be useful. IE on the opteron, 14-16 plies in the middlegame were normal, in endgames 21 plies and beyond are typical. But how to compare when one game enters the endgame at move 20 while another game doesn't reach the endgame until move 50... Even worse, if I retune search extensions I can increase or decrease the average ply reached at will, making comparison even harder... > > >also if in one game the average was 19.3 ply for example how much variation >would you expect between different games, on the same time and hardware. ? > Same game played multiple times? There would be some variation just based on timing changes, hash interaction, and parallel search non-determinism. For different games, the average ply depth could vary by a factor of 2 or more when one game reaches the endgame quickly and the next does not... > >duncan
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.