Author: Matthew Hull
Date: 16:06:30 08/22/02
Go up one level in this thread
On August 22, 2002 at 18:07:47, Uri Blass wrote: >On August 22, 2002 at 17:53:22, Uri Blass wrote: > >>On August 22, 2002 at 17:40:48, Matthew Hull wrote: >> >>>On August 22, 2002 at 17:13:35, Uri Blass wrote: >>> >>>>On August 22, 2002 at 16:39:38, Matthew Hull wrote: >>>> >>>>>On August 22, 2002 at 16:09:06, Chris Carson wrote: >>>>> >>>>>>On August 22, 2002 at 15:32:32, Matthew Hull wrote: >>>>>> >>>>>>>On August 22, 2002 at 15:15:08, Uri Blass wrote: >>>>>>> >>>>>>>>On August 22, 2002 at 14:37:07, Matthew Hull wrote: >>>>>>>> >>>>>>>>>On August 22, 2002 at 14:20:41, William H Rogers wrote: >>>>>>>>> >>>>>>>>>[snip] >>>>>>>>>>... That they beat the worlds champ in 2 out of 3 is on >>>>>>>>>>record, but that does not make them the worlds best. >>>>>>>>>>Bill >>>>>>>>> >>>>>>>>>It does if no other program has done it or can do it. I guess we'll soon know >>>>>>>>>the answer if the Fritz/Kramnik thing ever happens. >>>>>>>> >>>>>>>>Kramnik gets the machine before the match. >>>>>>>>The interesting match is kasparov-Junior and not Kramnik-Fritz. >>>>>>>> >>>>>>>>I think that you should learn statistics and 6 games are too little data to >>>>>>>>decide and stopping to play after these 6 games suggest that they know that they >>>>>>>>were lucky and they want to give the public the worng impression by not playing >>>>>>>>more games. >>>>>>>> >>>>>>>>I do not fall for this trap. >>>>>>>> >>>>>>>>Uri >>>>>>>>Uri >>>>>>> >>>>>>> >>>>>>>Let's put it this way. They are the only one's to ever get that lucky. But for >>>>>>>some people, beating the World Champ in 6 games with a lot of skill and some >>>>>>>luck (maybe even a lot of luck) adds up to zero. I think it adds up to > zero. >>>>>>>Can you at least admit that? >>>>>> >>>>>>Results for DB and Commercials: >>>>>> >>>>>> >>>>>> ELO Opp >>>>>>DB97 2862 2795 +2 =3 -1 >>>>>> >>>>>>Tiger 2788 2497 +8 =3 -0 >>>>>>DJ6 2702 2792 +2 =5 -2 >>>>>>Rebel 2697 2697 +2 =0 -2 >>>>>>DF6 2678 2545 +6 =4 -2 >>>>>> >>>>>>DB96 2642 2775 +1 =2 -3 >>>>>> >>>>>>Based on results, the top commercial programs are equal to DB today. >>>>>>The 2700 GM that Rebel tied with played 100 preparation games, GM Kasparov vs DB >>>>>>had none. >>>>> >>>>>How many preparation games did Andersson get before playing Morphy? Does anyone >>>>>dispute the fact that Morphy beat Andersson? Are there endless debates about >>>>>who was _really_ the stronger player? >>>> >>>>He played no preperation games against a copy of the opponent >>>>when the opponent does not know about it and this is the point. >>>> >>>>> >>>>>Was Fischer really stronger than Spassky? He only just barely beat him. He >>>>>probably would have lost if he hadn't made such a rediculous fuss and disrupted >>>>>the match. And he was too chicken to play Karpov. But I don't see you or >>>>>anybody else sticking up for Spassky or Karpov. >>>> >>>>Fisher played a lot of games. >>>>He beated some players 6-0. >>>> >>>>Fisher was better than spassky also based on public games >>>>against other players. >>>> >>>>It was not the case with deeper blue because it played no >>>>public games against other players. >>>> >>>>>>The commercial programs can be played by anyone. DB was only played >>>>>>by a select few team members and a few games later by DBjr. The commercial >>>>>>programs are the best today and results as good as DB and over more games. DB >>>>>>is a dead horse. >>>>> >>>>>According to the laws of arithmetic, 2862 is still greater than 2788. And it >>>>>was achieved against _THE_ WORLD CHAMPION. None of the others can boast that. >>>> >>>>Humans could not get experience against something similiar to the thing >>>>and the situation today is different. >>>> >>>>Uri >>> >>>From what I've read in these threads, especially the contributions from Dr. >>>Hyatt and his knowlege and conversations with the team members, it seems that >>>the DB2 team was caught off guard by their victory. They were expecting maybe >>>at best a draw and at worst, some well fought losses. They were expecting to do >>>one more improvement iteration. >>> >>>But when they won, IBM management stepped in and shut down the project and >>>reaped the marketing rewards. >>> >>>Well, give them a break then. Yeah, the logs weren't available, the re-match >>>didn't happen. Just a lot of stuff they hadn't planned for happened and didn't >>>happen. >>> >>>From the outside, I guess their behavior looked strange, rude, unaccountable. >>>But if you think about it (especially if you've ever worked in a >>>mega-corporation), it makes perfect sense. >>> >>>Why people can't see this is beyond me. Why do we need to hear you guys >>>discredit their work, belittle their playing strength, and pooh pooh their >>>victory? What is there to be learned from such stuff? >>> >>>All the other DT/DB machines were the best. DB2 was surely that and much more. >>> >>>Why is that so hard to accept? >>> >>>You will say "I analyzed the games. They didn't impress me". Yet they were >>>good enough to rattle and _defeat_ THE WORLD CHAMPION. >>> >>>Even THE WORLD CHAMPION HIMSELF was impressed with at least _one_ move. >>> >>>I'd bet GNUChess running on a 10ghz processor could beat today's commercials on >>>their 2ghz machines. Surely, DB2 would have an even greater speed advantage. >> >>I'd bet that GNUchess on 10gh cannot beat the commercial >>with 2Gh except maybe blitz and I doubt if it even can do it >>in blitz. >> >>Gnuchess even lost against my movei in a match of peter berger >>on equal hardware and movei is not close to the top programs. >> >>It is weaker than them in tactics and it has almost >>no knowledge in it's evaluation. >> >>Uri > >I can add that there was a big improvement in chess programs in the last year so >I can believe that gnuchess could beat the top programs of 1997 >with time advantage of 5:1. > >I read that gnuchess beated the top programs when it had >better hardware some years ago but software gets better and >I guess that today it is going to need >at least 1 hour per move against 3 minutes per move only to get 50%. > >Maybe Bob hyatt can try to play a winboard tournament of gnu >against Crafty when Crafty is using hardware is >5 times slower(if it is impossible than telling crafty to >do nothing in part of it's search can emulate this situation). > >Crafty is not the best program but I expect even Crafty >to get more than 50% when the time control is slow enough. > >Uri I will try it myself. I have an old 166mhz and a 900mhz Duron on my home network. GNUChess will get the Duron. I already did this test with Phalanx 22, which smashed Crafty on the 166mhz, which is not surprising. Phalanx is pretty strong anyway.
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.