Author: Derek Paquette
Date: 22:06:03 10/05/04
Go up one level in this thread
On October 05, 2004 at 22:28:58, Robert Hyatt wrote: >On October 05, 2004 at 20:15:56, Derek Paquette wrote: > >>On October 05, 2004 at 19:35:12, Robert Hyatt wrote: >> >>>On October 05, 2004 at 19:17:06, Derek Paquette wrote: >>> >>>>On October 05, 2004 at 17:12:17, Robert Hyatt wrote: >>>> >>>>>On October 05, 2004 at 14:14:13, Jorge Pichard wrote: >>>>> >>>>>>On October 05, 2004 at 11:40:27, Robert Hyatt wrote: >>>>>> >>>>>>>On October 05, 2004 at 03:25:32, Jouni Uski wrote: >>>>>>> >>>>>>>>Where is all the discussion?? >>>>>>>> >>>>>>>>There are different dates for games, but start date seems to be 8.10. And after >>>>>>>>4 days we know the truth about computers playing level. >>>>>>>> >>>>>>>>Jouni >>>>>>> >>>>>>>How will we know the "truth" after these games, when we apparently don't know >>>>>>>the "truth" after all the previous human/computer games??? >>>>>> >>>>>> >>>>>>What truth are we referring about? My only conclusion is that computers tactics >>>>>>are so strong nowadays that any strategic advantage that the human GM might have >>>>>>over them simply balanced out. >>>>>> >>>>>> >>>>>>Jorge >>>>> >>>>>I have no idea about what "truth" he was talking about. >>>>> >>>>>But the computers are _not_ overwhelming the humans in tactics by any stretch... >>>> >>>>YOu can look at that two ways, >>>>1. computers aren't creating or isolating tactical situations and exploiting >>>>them >>>>2. computers will play tactics near perfectly every time, where a human would >>>>not. So you could almost say they are blowing humans off the board with >>>>tactics. >>> >>>I wouldn't say any such thing. Give a computer Shirov's Bh3 sacrifice and see >>>how long the "tactical monsters" take to see that, and it is _all_ tactics. The >>>main advantage of computers is steady play. Humans occasionally make _big_ >>>mistakes. Computers simply do not. Apparently that is enough to produce pretty >>>good results... >> >> >> pretty good is an understatment in my opinion. How many times have we seen >>computers finish first in a tournament? And these same computers runing these >>programs (noteably shredder 8) can be bought off the shelf. So by saying 'good' >>is a understatment. Infact the book used in Argentina was the same book from >>the box. >> >>CT15 in Argentina finished first, people didnt know its style as much and got >>blown away. >> >>This upcoming small tournament will be a good test to see just how well 'steady' >>play assists in elo points vs humans. >> >>Yes there are certain moves that computers can't find, but there are a lot of >>moves overlooked in tournaments by humans because there are simply too many >>things to consider in a 3 minute per move time frame, so they aren't blunders >>persay, they are just the runoffs of the advantages of a computer over a human. >> >>So i agree with you that steady play is one of the reasons, but I disagree with >>your 'pretty good' results. >> >>-Derek Paquette > > >In matches against humans, _what_ program has won one of those matches, with the >sole exception of Deep Blue vs Kasparov in 1997? > >And the moves I recall _are_ blunders. Not in overly complex positions either. >Just plain and simple blunders... the last kasparov vs comp match had a couple >of good examples... Kasparov and kramnik both drawed their programs, Hydra just recently beat a GM, with 3 wins and 1 draw (can't remember his name) Matches are not necessarily as important as tournaments either for judging machine strength. Much like a simple match between humans or a large tournament. If player X can beat player Y 80% of the time, but X loses to all other players 50% of the time where as Y beats all other players 60% of the time, and there are twenty other players, Y is the clearly the stronger player, So if you throw a bunch of humans into a tournament with a program, that is a much better test than just matches. Shredder 8 and CT14 both came first in their respected tournaments.
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.