Author: Robert Hyatt
Date: 19:08:07 10/02/98
Go up one level in this thread
On October 02, 1998 at 20:38:02, James Robertson wrote: >Whenever I ask programs like Crafty or Chessmaster to display their principal >variations, if they search like ply 6, they display 6 plies of pv plus >extensions. I.e, it looks something like this: QxP Rc7 Bd6 Rb7 Qd2 Rd1 Qf2. But >in other programs (Arasan, Exchess) if they say they searched 6 plies, the pv is >often only 2 or 3 plies long: QxP Rc7. I'm guessing this is because of hash >tables ? Anyway, no matter how they come up with a pv, how does Crafty avoid >displaying a pv much shorter than the claimed number of plies searched? >Conversely, how does Exchess decide to only display a portion of the plies >searched? > >Thanks, > >James I do two things... one, PV's can end with a <HT> which is where they were cut short by a hash hit. But when this happens, I walk the PV to the end, then do another hash probe to see if there is a suggested best move. If so, I add this to the PV and then probe again in the same way, trying to recover the part (or some part of the part) of the PV that was cut short... Other programs probably do different pruning algorithms, which might well let a N ply search quit "early"... Crafty always searches down the PV to the depth it claims, plus some for extensions...
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.