Author: Robert Hyatt
Date: 10:14:48 10/28/02
Go up one level in this thread
On October 28, 2002 at 12:45:11, Christophe Theron wrote: >On October 28, 2002 at 05:28:39, José Carlos wrote: > >>On October 28, 2002 at 05:02:13, Nagendra Singh Tomar wrote: >> >>>On October 28, 2002 at 04:52:26, Uri Blass wrote: >>> >>>>On October 28, 2002 at 04:24:24, Nagendra Singh Tomar wrote: >>>> >>>>>On October 28, 2002 at 02:45:33, Uri Blass wrote: >>>>> >>>>>>On October 28, 2002 at 02:00:06, Nagendra Singh Tomar wrote: >>>>>> >>>>>>>On October 28, 2002 at 01:48:02, Ricardo Gibert wrote: >>>>>>> >>>>>>>>On October 28, 2002 at 01:17:41, Nagendra Singh Tomar wrote: >>>>>>>> >>>>>>>>>What all reasons can you think of for a ply 'n+1' searching lesser nodes than >>>>>>>>>ply 'n' search. >>>>>>>>> >>>>>>>>>regds >>>>>>>>>tomar >>>>>>>> >>>>>>>>The most obvious is a ply n search had really bad move ordering that gets much >>>>>>>>improved for ply n+1 thanks to the hash moves saved in the hash table. >>>>>>> >>>>>>>correct. >>>>>>> >>>>>>>One more reason can ne that 'n' was just small enough to deny null-move. >>>>>>>null-move in 'n+1' ply search saved a lot of nodes. >>>>>> >>>>>>I use null move from the first ply so there is not n that is small enough to >>>>>>deny null move. >>>>>> >>>>>>The only case when I do not start by checking null move is when the remaining >>>>>>depth is 0 or in some other special cases when I suspect a zugzwang. >>>>>> >>>>>>Uri >>>>> >>>>>But the idea behind null move is to try first with a lesser depth search and see >>>>>if it fails high .. >>>>>if you start null move search from first ply .. what is the "lesser depth" you >>>>>are going to try for ply 1 and even 2 >>>>> >>>>>regds >>>>>tomar >>>> >>>>depth 0 is less than 1 or 2. >>>> >>>>Uri >>> >>>Do you really believe depth 0's result so much as to trust it for a fail high. >>> >>>tomar >> >> Sure. If null move reduction (R) = 2, which means you substract 3 to nominal >>depth and search for the same side, while your nominal depth is <= 3 null move >>will go directly to QSearch (depth 0). >> >> José C. > > > >You can't use that in the first iterations because QSearch is not guaranteed to >return a best move at all. It returns a move only if there exist a non-losing >capture. > >If the first iterations, null move is not doing anything because of this. > > > > Christophe It's still ok. Q-search returns a "stand-pat score" if there is no capture. That still makes the search work just fine...
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.