Author: Dann Corbit
Date: 12:57:36 01/29/99
Go up one level in this thread
On January 29, 1999 at 15:49:18, Matt Frank wrote: >It is my position that we have now moved clearly into a dimension where software >obtained for under $100 dollars can play at grandmaster level while played on >hardware obtained for less than $2000. For example, it is my contention that my >G3 233 Power Mac purchased in early '98 for much less than $2000, is a 2600 + >rated elo performer (2610 give or take a few dozen points) using MAC Hiarcs 7. >Other software, Fritz 5.32, Rebel 10 EOC--anti-grandmaster, and Chess Tiger 11 >+, among others should be similarly endowed if they could play on my MAC. These >other programs played on P2s with 400 MHZ processors should have similar or >better results. On what basis do you make this assumption? Computers have not been tested extensively against humans under tournament conditions. The SSDF ratings are computed against other computers, for instance. Certainly the computers are very strong. But are they GM level? And the sneaky radio deal does not convince me either. >The key question is this, when are we going to sweeten the pot sufficiently >(i.e., cash prizes) so that Grandmasters rated FIDE between 2500-2650, will find >it attractive to compete against our demons? A match like this would clearly >demonstrate once and for all the present level of these hardware software >configs. I know all about Rebel 9 against Yusupov, and Rebel >10--EOC--Anti-grandmaster against Anand. Yusupov is a 2650 player, and Anand is >just shy of 2800 elo FIDE. In either case the human won under tournament >conditions (40/2hrs.), which is the best playing speed for humans vs computers. Who is "we"? Are you going to supply the prize money? If a computer wins a tournament, does it get the prize? If you play against a computer and cannot benefit and can only lose out, would you want to do it? >I think that a 6 or 8 game match against a Grandmaster of decent reputation >(e.g., Christiansen, Benjamin, Yermolinsly, de Firmian, etc) would be a strong >test. I believe that my machine would not be at a disadvantage against any of >these players at 40/2hrs. Of course quicker timed games would be to my systems' >advantage. Therefore, the fairest test for a human against Hiarcs 7 G3 233 would >be games at least all moves in 25 minutes or longer. I would love to see a match like this. What would be really interesting is 32 GMs against the 32 top programs in a week long tournament. That would give us some really excellent data. >Are there matches between human grandmasters and the latest software running on >machines as fast as I have poroposed? Aside from the Deen Hergott vs Hiarcs 6 >match (won by hiarcs 6 4-2 on a P2 200 64 Ram machine in 1997) I know of no >other matches. Hergott is rated 2370 FIDE elo and is an International Master. If >you assume that Hiarcs 6 on that machine was rated FIDE 2525 or so based on the >SSDF ratings then the score was well within expectations (i.e., a score of 4-2 >would be consistent with an elo diff of about 120 points, where in fact it is >about 155 if you take the SSDF rankings as reasonable facsimiles of the FIDE elo >ratings. Incidentially, when Hergott played this match he was actually rated at >2495 FIDE elo based on his 1996 ratings. GM Matches against computers are very, very rare under formal conditions. They happen all the time on ICC and FICS. >My question is, can I get comments about how you all feel about these prospects? >I'd prefer that we don't quibble about the ratings (even though I am sure many >of you have serious doubts about my assumptions--I'd love to play this match to >prove my point), rather, that we focus on how such a match may be culminated. The GM's have nothing to gain and everything to lose. Why should they want to play? And if we are going to come up with $100,000 prize money, who is going to cough it up?
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.