Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How many killers? & Random chess & Has anyone NOT tried this?

Author: Robert Hyatt

Date: 11:42:13 09/27/98

Go up one level in this thread


On September 27, 1998 at 09:36:12, Jari Huikari wrote:

>On September 24, 1998 at 06:19:18, Roberto Waldteufel wrote:
>
>>Searching captures first is indeed very beneficial for alpha-beta efficiency,
>>but I think that just about all other programs also do this, as it is well
>>known. By the way, you should include promotions as well, whenever there is
>>one available, as these too cause a big change in material.
>
>(I)
>What about advancing white pawn to seventh (black pawn to second) rank?
>Should also be examined early?

no. that's the purpose of "killers".  If the pawn move turns out to be
good, it will become a killer and tried early from then on anyway...  and
when it just tosses the pawn, it won't be a killer and won't waste time.

>
>(II)
>HOW MANY killer-candidates your programs store? (To be tried if they cause
>a cutoff.)



I've always used 2 per ply.  In Cray Blitz, since I didn't use the history
heuristic, I also did some other killer tricks.  IE try both killers for
the current ply, they try the killers from ply-2, and even ply-4...



>
>(III)
>I added random feature to Nero 4. (Explained in the end of its FAQ-file.)
>Makes randomly one of the legal moves. Can OCCASIONALLY beat KASPAROV... :-)
>
>Latest version of Nero  http://www.math.jyu.fi/~huikari/NERO4.ZIP
>
>					Jari



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.