Author: Inmann Werner
Date: 16:16:39 11/20/98
Go up one level in this thread
On November 20, 1998 at 15:02:23, Will Singleton wrote: > >On November 20, 1998 at 11:03:09, Inmann Werner wrote: > >>On November 20, 1998 at 09:35:46, Robert Hyatt wrote: >> >>>On November 20, 1998 at 06:28:01, Inmann Werner wrote: >>> >>>> >>>>Do you really make null moves in the PV. How do you manage, that Null moves do >>>>not get into the PV? >>>> >>>>Werner >>> >>> >>>Because when we do the null-move search, if it fails high we return beta >>>which means it *can't* get into the pv (failed high). If it doesn't fail >>>high, we ignore it and do a normal search, which means something else will >>>get into the pv with the null-move being ignored totally. >> >>Thanks a lot, Robert and Peter!!!! >> >>This was one of the most simple and good tips, I got!!! >>I now use null move also in the PV, had to change only one variable in my >>program and my search tree is reduced by 40%!!!! (tested with 4 positions) >>It must be some brain-damage of me, why I did not recognize this of my own. >>There you go looking for speed increasments of 2 to 5 percent, and here you get >>a lot of this as simple! >> >>Thanks again >> >>Werner > > >40 percent? Seems like a lot. You might want to check your results on the full >WAC. I would think you'd solve a lot more with the added depth, and if not, >then you've got a problem. I'd be interested in knowing. > >Will I tried some normal Games against a strong program, and came most of the time one depth deeper. There were no "bad behavior" or so. Also in the testpositions of LCTII I found nothing strange. The program used the same PVs and rates. I used null move before, but not in the PV. Because the PV is the most tree using thing, the reduce is ok. In LCTII positions, the 40% are right. Also the programs solves the same positions as before in LCTII, but faster. Till now I found nothing bad, but I will test further. I also think, that the 40% are not normal for all programs. My program before really was very slow at PV nodes. Werner
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.