Author: Chris Carson
Date: 06:16:16 01/06/00
Go up one level in this thread
On January 06, 2000 at 06:00:12, Graham Laight wrote: >On January 05, 2000 at 18:50:42, Peter Kappler wrote: > >>On January 05, 2000 at 12:58:55, Graham Laight wrote: >> >>>On January 05, 2000 at 11:52:28, Peter Kappler wrote: >>> >>>>On January 05, 2000 at 04:51:43, Graham Laight wrote: >>>> >>>>>On January 05, 2000 at 01:05:20, Peter Kappler wrote: >>>>> >>>>>>I'd still take Kasparov or Anand in a G/30 match against any micro, though it >>>>>>would certainly be competitive. >>>>> >>>>>You mean like Kasparov did in London, in 1994, against Genius 3 on a Pentium 90? >>>>> >>>>>-g >>>>> >>>> >>>> >>>>I'd want the match to be longer than 2 games, so accidents like that don't >>>>influence the result too heavily. Make it 24 games and I'd feel pretty good >>>>about GK's chances. >>>> >>>>--Peter >>> >>>Hi Peter, >>> >>>Didn't 2 other GMs have the same "accident" in the same tournament? >>> >> >>I don't know what tournament you're talking about. My recollection is that the >>games versus Genius were isolated exhibition games, and not part of any >>tournament. If you have details, please post them. > >In 1994, Intel sponsored a tournament in London - 15 GMs and one computer at >active time controls. I think that Anand managed to knock out the computer in >the end. > >>>Isn't 2 games enough for a GM to get at least one win against Genius 3 on a >>>Pentium 90 - even at active time controls? >>> >> >>Maybe, maybe not. My point was that if you're trying to accurately judge the >>computer's playing strength, you need to see more than 2 games. >> >> >>>When we discuss computer v human strength, I think that these "accidents" should >>>be taken into consideration - especially when the evidence being put forward by >>>the "computers are not yet even 2500" brigade seems to be mostly of a similar >>>anecdotal nature. >>> >> >>I regret using the work "accident" to describe that game. Kasparov certainly >>deserved to lose it - my point was just that I think Garry would have finished >>well ahead of Genius in a longer match. > >Agreed! > >>As you said, this whole debate is largely anecdotal because there aren't enough >>games to draw firm conclusions. I have stated many times that I think computers >>are probably in the 2450-2500 range, but that we really need more 40/2 comp vs. >>human games to draw conclusions. >> >>I believe that if you take all of the recent Rebel games and compute a >>performance rating, you'll find that it is very close to 2500. (It was 2485 >>before the 2 games against IM Russek). This is pretty clearly below the >>requirements for a GM norm. > >I thought Russek was a GM, not an IM? > >Anyway, Century does not have an SSDF rating. The last Rebel rating is: > >15 Rebel 9.0 47MB P200 MMX 2522 24 -23 900 61% 2443 > >Assuming that, although Rebel now plays more stylish chess, it's actual rating >hasn't improved (which is credible - the same thing happened to Genius from V3 >to V5), then it's FIDE rating would be about 2492. Give it, say, 40 extra points >for running on faster hardware, and the result is a score of 2532. The margin >for error (95% confidence) is 23, so its score could be 2509 - which is about >what Ed is claiming for Century in CCC. > >So you see - the Rebel GM challenge is hardly weakening my position! > >-g Your argument is very strong. 95% confidence is 2 errors of measure, so 2532-23-23 = 2486 and a 99.7% confidence is 3 errors of measure, so 2532-23-23-23 = 2453. Rebel has performed well in match play and in my opinion would do better in tournamets (where participants would not spend a lot of extra time in preparation to exploit Rebel Century weakness). Also, if the FIDE event had players with an average rating of 2443 (which would match the Rebel opponets average score above). I think Rebel's performance rating would be higer in that tournament than in the matches it has played. :) IMHO. :) Best Regards, Chris Carson
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.