Author: Robert Hyatt
Date: 14:40:17 11/24/03
Go up one level in this thread
On November 24, 2003 at 16:38:25, Uri Blass wrote: >On November 24, 2003 at 16:19:19, Robert Hyatt wrote: > >>On November 24, 2003 at 11:10:42, Uri Blass wrote: >> >>>On November 24, 2003 at 09:07:39, Robert Hyatt wrote: >>> >>>>On November 24, 2003 at 07:41:16, Daniel Clausen wrote: >>>> >>>>>On November 23, 2003 at 17:29:28, Peter Berger wrote: >>>>> >>>>>>For information about setup and rules: >>>>>>http://www.talkchess.com/forums/1/message.html?329237 >>>>> >>>>>I quote from there: >>>>> >>>>>"Saying that the faster computer is about 10 times faster shouldn't be too >>>>>wrong." >>>>>and >>>>>"With this setup Crafty should be the clear favourite I suppose." >>>>> >>>>>I don't see why Crafty should be the clear favourite here. When I look at the >>>>>SSDF list, I see the following ratings: >>>>> >>>>> >>>>>Junior 8.0 256MB Athlon 1200 MHz 2784 >>>>>Crafty 18.12/CB 256MB Athlon 1200 MHz 2615 >>>>> >>>>>So there's a rating difference of 169 points. A factor of 10 is slightly more >>>>>than 3 times doubling and if we use 70 points for each doubling (which is the >>>>>upper limit in my opinion) we get the difference (3*70)-169=41 points. >>>>> >>>>>In a match where the first one to reach 6 wins is the winner, that's hardly >>>>>"clear favourite". It's more like using a dice to decide something. Or am I >>>>>missing something? >>>>> >>>>>Sargon >>>> >>>>Yes, he is not using 18.12. He is using something two years newer. >>> >>>newer does not mean better in comp-comp games based on experience with Crafty. >>> >>> >>>Crafty16.19 is the leader of the crafties of the following list >>> >>>http://www.digichess.gr/infiniteloop/ratings/NIL2_final.txt >> >>I can _guarantee_ you that 16.19 is not better than 19.5, at normal >>time controls. I can guarantee you it is significantly worse, in fact. >>For blitz, I don't know, but then again I don't really care either since >>we aren't playing blitz matches that count for anything. > >19.5 was not tested for that list but the time control was 80+3 on P3 1Ghz(no >pondering) and 16.19 did better than newer versions(I know also that you do not >like ponder off but I do not believe that it helped 16.19 relative to other >versions). > >What tests did you do to be sure that 19.5 is better than 16.19 at normal time >control(Note that I am talking only about comp-comp games and not about >comp-human games) I know what changes were made and what bugs were fixed. 19.x has much more endgame knowledge than 16.19, not to mention a better null-move search and other things... > > >> >>> >>> >>>It does not prove that Crafty16.19 is really better(not enough games) but it >>>gives serious doubt if newer Crafty is really better. >>> >>>I also remember that Crafty16.xx was the Crafty that beated Rebel in 1 hour per >>>move. >>> >>>It seems that the commercial programs did big step forward from that time when >>>Crafty did not do it. >> >>I haven't seen my results get worse on ICC over the past N years. Things have >>(IMHO) remained fairly "steady-state". Occasionally there is a need for some >>minor tuning (IE when Gambit Tiger hit the streets) but that was done by many >>relatively quickly to restore status-quo. > >Note that I thought only about games against computers and ICC includes games >against humans. > >Another question to check is if your hardware did not improve more than the >hardware of the opponents. > >Uri Nope. My xeon 2.8 is not any faster than many machines running on ICC, and it is actually slower than quite a few. There are several dual 3.06 and 3.2ghz xeons running, not to mention AMD athlon duals...
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.