Author: Robert Hyatt
Date: 10:25:11 01/28/98
Go up one level in this thread
On January 28, 1998 at 12:30:15, Stuart Cracraft wrote: >I am curious what others are using for a >move selection order in the full-width >and quiescence parts of their program's search. > >Mine has the current following order for full-width > > move from the hash table > captures based on MVV/LVA > killer moves > history heuristic > centrality I handle captures differently. I look at winning captures ordered by my Swap() function, then even exchanges as confirmed by the Swap() function, but I stop there. I defer the losing captures (those not chosen by the two above steps) until after killers and history moves. However, I get a 10% improvement in tree size when using Swap as opposed to MVV/LVA with no other changes. But using MVV/LVA means you can't easily recognize losing captures and defer them until after the killers and so forth. > >For quiescence > > captures based on swap-off routine, ties broken by MVV/LVA > promotions based on promoted piece value I simply use Swap and ignore ties. The extra tests you do might actually hurt. > >Notes: > - Doesn't search checks in the quiescence unless the move > happens to be a capture or promotion. > - Doesn't use swap-off to order captures in full-width due to > tests not producing any improvement > >Thanks for any comments. > >--Stuart
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.