Author: Robert Hyatt
Date: 20:23:19 08/01/02
Go up one level in this thread
On August 01, 2002 at 18:10:26, Dieter Buerssner wrote: >On August 01, 2002 at 18:02:36, Ulrich Tuerke wrote: > >>On August 01, 2002 at 17:45:33, Dieter Buerssner wrote: >>>If you get a fail high here - why research? You already know that it is not an >>>"easy move". >> >>... because it could be even better than the current best move. > >Ooops. Of course - thanks for pointing out my blindness. But still, in this >case, wone could speed it up a bit. Only research with window(s-1,s) and only >when this fail high, too, research with a bigger beta. Exactly the same as a >normal PVS search at the root would do. So, I still think the fail high case >here is not special, and the added effort comes not after the first fail high >with the lower window, because everything what follows, would need to be done in >a normal search anyway. Or is there some other blindness in my argumentation? > >Regards, >Dieter Remember, that in normal PVS, if a move after the first fails high, it is _known_ to be better than the best so far. In this offset window search, any move can fail high on the offset window but fail low on the normal window. And that fail low (a) is not free; and (b) would not occur in a normal non-offset search...
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.