Author: Mark Young
Date: 18:55:12 06/22/99
Go up one level in this thread
On June 22, 1999 at 21:32:06, James T. Walker wrote: >Hello, >I believe the mystery of why Hiarcs 7.32 plays better in engine/engine games is >because of the way Hiarcs does not reset it's hash tables after each move. This >is an advantage because Fritz does. The result is that Fritz must start it's >search at ply 1 each time even if it has predicted Hiarcs move because no >pondering is allowed in engine/engine testing. But because of the way Hiarcs >uses it's hash when it is Hiarcs turn to move it still has search info in the >hash memory. So you will see Hiarcs start it's search at something like: >+-2 (2.68) depth 7/18 00:00:00 27kN. This is almost as good as pondering >because it saves time in the search. Effectively Fritz is stripped of it's >"Pondering" while Hiarcs has something almost as good as pondering. Note this >advantage does not exist in Auto232 operation since Fritz is pondering and when >it guesses Hiarcs next move it is effectively way ahead in it's search. > >Jim Walker > >P.S. Any comments welcome. Possible, If so I would expect the advantage to be more pronounced at faster time controls. As the chance of missing tactics is greater the faster the time control. And any kind of search advantage should show greater in the results. I have only done a few games at 10 sec or slower, but my engine vs engine results agreed with your data.
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.