Author: Sune Fischer
Date: 02:01:56 04/25/02
Go up one level in this thread
On April 25, 2002 at 04:54:39, Uri Blass wrote: >On April 25, 2002 at 04:44:15, Sune Fischer wrote: > >>On April 24, 2002 at 18:42:25, Uri Blass wrote: >>>I can add that I was not sure if recursive null move is really productive at >>>long time control inspite of the fact that it is productive at blitz based on my >>>tests of many blitz games so I tested it against chezzz at 400 minutes per >>>game(chezzz is a program of the 3th division) and movei produced the following >>>game(note that chezzz beated a previous version of movei at the same time >>>control): >>> >>>I did not tell instructions for chezzz about hash tables but I know that chezzz >>>was using 8 Mbytes hash tables as default value when movei still does not use >>>hash tables. >>> >> >>You should give Chezzz a bit more than 8 MB hash, if you give it 64 MB or 128 MB >>it will be stronger, especially for those long time controls. >> >>-S. > >I read that doubling the hash tables give only 6-7 elo improvement. >It means that 128Mbytes instead of 8 Mbytes are going to give Chezzz only 24-28 >elo improvement. > >I also wanted to use the same conditions that were used for the previous >version. > >Uri I know the hitrate goes up quite a bit, and I don't believe that Chezzz was intended to run with only 8 MBs of hash. It is just a low default for people that doesn't have a lot of system memory. In all fairness, chezzz should not be punished like that just because movei hasn't got a hash, even if it is "only" 25 elo (I suspect it is more). -S.
This page took 0.01 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.