Author: Robert Hyatt
Date: 07:13:25 09/21/02
Go up one level in this thread
On September 21, 2002 at 02:22:16, Terry Ripple wrote: >On September 20, 2002 at 16:56:12, Robert Hyatt wrote: > >>On September 20, 2002 at 15:45:45, Terry Ripple wrote: >> >>>On September 20, 2002 at 10:36:11, Robert Hyatt wrote: >>> >>>>On September 20, 2002 at 08:12:17, Chris Carson wrote: >>>> >>>>>On September 19, 2002 at 23:10:11, Robert Hyatt wrote: >>>>> >>>>>>On September 19, 2002 at 21:55:44, Rick Terry wrote: >>>>>> >>>>>>>On September 19, 2002 at 17:34:04, Robert Hyatt wrote: >>>>>>> >>>>>>>>On September 19, 2002 at 14:34:48, Chris Carson wrote: >>>>>>>> >>>>>>>>>On September 19, 2002 at 14:17:50, Robert Hyatt wrote: >>>>>>>>> >>>>>>>>>>On September 19, 2002 at 14:05:58, Jorge Pichard wrote: >>>>>>>>>> >>>>>>>>>>>On September 19, 2002 at 14:03:25, Jorge Pichard wrote: >>>>>>>>>>> >>>>>>>>>>>If Hiarcs 8 can at least use the new AMD 2.6 Ghz or the upcoming 3.0 Ghz Intel >>>>>>>>>>>by January, it might have a chance to win. >>>>>>>>>>> >>>>>>>>>>>http://www.chessevents.nl/bareev_match.shtml >>>>>>>>>> >>>>>>>>>> >>>>>>>>>>Note that this is 40 moves in 2 hours. >>>>>>>>>> >>>>>>>>>>The computer is going to have its "hands" full with this GM. Or any >>>>>>>>>>GM. At that time control. >>>>>>>>> >>>>>>>>>At 40/2 on AMD 2.6 or PIV 3.0, the advantage is to the comp. The GM may win, >>>>>>>>>but Hiarcs 8.0 is the favorite in this match. >>>>>>>>> >>>>>>>>>This is great news, my best to both the GM and Hiarcs. >>>>>>>> >>>>>>>> >>>>>>>>I totally disagree. The longer the time control, the better the human will >>>>>>>>do. Based on watching these games about 30 years now. IE at correspondence, >>>>>>>>a good IM will tear the chips out of most any program going... At blitz, the >>>>>>>>comp is nearly unbeatable... _nearly_ being the operative word of course. >>>>>>>> >>>>>>>>Hiarcs might be the favorite for you, but my instinct says "human". At game/30 >>>>>>>>the comp would definitely be favored. At game/60 it gets tougher. At a non- >>>>>>>>sudden-death time control, the human isn't going to get into time trouble and >>>>>>>>get blitzed... >>>>>>> >>>>>>> >>>>>>> Well Century 4 seemed to handle Van Wely quite easily, Hiarcs is much stronger >>>>>>>then Rebel on faster hardware. >>>>>> >>>>>> >>>>>>I don't think it "handled him quite easily"... >>>>>> >>>>>>but we will see before long,.... >>>>> >>>>>Feb, 2002 Rebel Century 4 on an AMD 1900+ plays an even match against GM Van >>>>>Wely (Fide 2697, top 10 GM at the time) at 40/2. Score for Rebel is +2, 0, >>>>>-2(no draws, all wins for both players were with the white pieces). >>>>> >>>>>It was an even match and on slower HW. I would agree the in this match the GM >>>>>and computer are about even (but I like the comps chances if on 3Ghz machine, >>>>>with programmer operating and latest s/w). >>>>> >>>>>I agree with one of your later posts that most GM's can play the comps close, >>>>>but the advantage is now with the comps against most GM's at 40/2. Perhaps 40/3 >>>>>would be better for the human GM's. A top 10 GM is about even on 2Ghz at 40/2, >>>>>this will be a good match, in 18 months it will need to be a Top 5 GM at 40/2. >>>> >>>> >>>>I think you are _greatly_ over-estimating the mhz contribution. In comp vs >>>>comp, we pretty well know what additional nps will do. But the same does not >>>>appear to carry over to comp vs gm. As has been seen on ICC many times. I >>>>don't remember the details now, but several of us ran some tests on ICC a >>>>few years ago, showing that doubling the cpu speed had no real effect on >>>>overall score against GM players. I ran the test with Scrappy, running it on >>>>a laptop at 1/4 the speed of the normal machine I was using at the time, and >>>>there was very little difference in overall results against the same players >>>>(roman, udav, yasser, christiansen, etc... >>>> >>>>I won't say there was _no_ difference, but there was no 100+ rating change >>>>either, not even close... >>>----------------------------- >>> >>>Hi Bob, >>> >>>Is it possible that because you were playing "blitz" time controls that this was >>>the reason for not seeing any or little difference in the results compared to a >>>program playing at 40/2hr? >>> >>>Terry >> >> >>I don't know. I do know that at one point, I had a quad P6/200, and my quad >>xeon 400, and I swapped them during testing, and during one marathon with Roman, >>playing 5 12 games, I asked "can you tell the difference between crafty at 2pm >>today and crafty now? and he said "not that I can tell." One was well over 2x >>faster than the other. >> >>I also ran them "side by side" and there was little difference against the >>GM players, although there was a noticable difference against other comps... >> >>Not very scientific, however. But it makes sense to me. Yes they get better >>as they get faster, but if the games are not tactical, deeper searching doesn't >>improve positional play nearly as much as it improves tactics. >---------------------- > >Well, this makes sense! The games that were played during this test, did you >ever check what percentage were "tactical" vrs. "positional"? > >Terry In the case of Roman, games don't tend to get very tactical, very often. Because he knows what is going on and steers the game away from tactics and into positions where his "understanding" is more important than sheer calculation. If a GM wants to "slug it out" then the games (and results) are far different as speed becomes more important...
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.