Author: Uri Blass
Date: 22:40:14 06/13/01
Go up one level in this thread
On June 14, 2001 at 00:21:39, James T. Walker wrote: >On June 13, 2001 at 18:23:03, Uri Blass wrote: > >>On June 13, 2001 at 17:40:07, James T. Walker wrote: >> >>>On June 13, 2001 at 16:05:55, Uri Blass wrote: >>> >>>>On June 13, 2001 at 15:59:04, Peter Ackermann wrote: >>>> >>>>>If I compare the March-01-edition and the June-01-edition I see that not a >>>>>single game was played with Hiarcs 7.32 for three month now. Both lists show >>>>>Hiarcs 7.32 with 679 games. >>>>>Is Hiarcs 7.32 not attractive anymore? I myself consider Hiarcs 7.32. still one >>>>>of the very interesting engines and if I try engine-matches between the Tigers >>>>>ans Hiarcs it is not so a clear result..... I have the impression that Gambit >>>>>Tiger has some problems with the style of Hiarcs 7.32. So please give Hiarcs >>>>>7.32 a chance from time to time even if Hiarcs 8 seems to be on the way. >>>>> >>>>>Peter >>>> >>>>What is your hardware and what is the time control of your game? >>>> >>>>It is important because Hiarcs7.32 does not like the K6-450 of the ssdf and I >>>>also believe that it prefers also fast time control and not 2 hours/40 moves. >>>> >>>>Uri >>> >>>Hello Uri, >>>I have seen post by you before like this and I don't agree. I would like to >>>know where you get your information. >> >>The fact that hiarcs does not like the ssdf hardware is known. >>I was careful to say "I believe" so I may be wrong and it is based on my >>impression from analyzing some engine-engine games and also on the fact that >>I found from my experience that hiarcs has a bad branching factor at long time >>control. >> >>Another reason that expect Hiarcs to suffer at long time control is the fact >>that hiarcs cannot search more than 30-32 plies forward when the time control is >>not important(I understand that it is going to be fixed in Hiarcs8). >> >> I'm still running Hiarcs 7.32 on my Athlon >>>900 mhz machines and it still does very well against other top programs. For >>>instance in my "standard" database (G/60 or longer) Hiarcs 7.32 is right now >>>ranked tied for 6 place with Chess Tiger 13.0. It is behind Gambit Tiger >>>14.0,Chess Tiger 13.0, Deep Shredder and Fritz 6e/6b. This puts it ahead of >>>programs like Gambit Tiger 1.0, Shredder 5, CM8K(32M), Junior 6.0 and Rebel >>>Century 3.0. This is not a huge database. It has only 694 games but so far I >>>see nothing to indicate Hiarcs does not compete with the faster hardware/longer >>>time controls. >> >>This data does not contradict or prove the conjecture that hiarcs is better in >>blitz on the same hardware. >> >>The only way to test if it is worse at long time control is by comparing >>hiarcs7.32's result in blitz with the results of it at longer time control on >>the same hardware. >> >>Using the same hardware is important. >> >>The reason is that I believe that hiarcs likes the fast hardware more than other >>programs(my experience is that for Hiarcs PIII800 is more than 2 times faster >>than pIII450 when it is not the case for Deep Fritz). >> >>Uri > >Well I'm still not sure what you are saying. I know that on my K6-3-450 Hiarcs >was the best program on the planet at G/1 minute or G/2 minutes but at >Game/5minutes Fritz was about even or maybe a little better. Now on an Athlon >900 I would guess that Hiarcs is good at G/30 seconds or G/1 minute now but at >G/5 minutes Fritz is definitely better since this would translate to roughly >G/10 minutes on the 450. You may be right here but you used the K6-3-450 when the ssdf used k6-2-450 My point is that Hiarcs may be 3 times faster on Athlon relative to the ssdf hardware when Fritz is only 2 times faster relative to the ssdf hardware. This seemed to be because of something special that >Hiarcs did which made it very fast compared to other programs in short time >control matches but offered no advantage in long time control games. In spite >of any branching factor problem it still competes very well on fast hardware and >long time controls. Maybe it's not so good at overnight analysis and I'm not >sure if any programs go 32 plies except in the very late endgame where the >tablebases should dominate the search anyway. >Jim There are cases in the middle game when seeing more than 32 plies in the selective search may be important. I can see often that the selective search of Fritz is more than 32 plies even in tournament games and I believe that the selective search of Hiarcs is not more than 32 plies(I am not sure about it and I may be wrong about it). Uri
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.