Author: Robert Hyatt
Date: 18:55:23 11/11/98
Go up one level in this thread
On November 11, 1998 at 12:40:37, Don Dailey wrote: >On November 10, 1998 at 14:50:07, Reynolds Takata wrote: > >>Is I.M. Kaufman, and the recent CCC poll on Comp strength the definitive word on >>whether comps are GM strength? I was just wondering because it AMAZES me that >>after so much debate on the subject that an International Chess Master states in >>"Computer Chess Web Reports" that he believes that a program "Mchess 8.1" would >>be able to acquire the 3 grandmaster norms in 40/2 tournaments to acquire the GM >>title, and that almost no one in the group commented on it. It's stunning this >>silence, especially since most CCC members will probably expect other programs >>(i.e. Rebel10, Fritz, Hiarcs etc.) to be stronger than Mchess. Of course I know >>it's not definitive but does this mean that the tide is turning hard against the >>ney sayers? Yes i know you might be able to find some other titled players to >>counter, though I don't know if they know as much about comp chess as Kaufman. >>Also every time i have asked titled players (5 players) on ICC if Comps could >>get the norms at 40/2 they all said yes. Though some didn't think comps could >>stand up to a lot of GM's in a match but they thought a tournament situation was >>different. Though after watching R10 draw anand in a 40/2 game i'm not certain >>even of that. >> >>Reynolds Takata >>USCF Life Master >>Fide Master >> >> >> >>"The last time I checked, the indicated adjustment to convert to USCF >>ratings was 180, which would give it a predicted USCF rating of 2696. Even a >>tiny gain would bring the new version over >>2700, the same level as the top four or five American grandmasters. This is at >>standard tournament speed (40/2); at fast >>chess the programs are stronger still. Although I suspect that the method used >>to get the Swedish ratings tends to exaggerate >>the ratings of the top programs a bit for several reasons, >> >>I do believe that MChess Pro 8.0 would be able to earn the >>Grandmaster title if given a reasonable number of opportunities to play in GM >>norm tournaments. >> >> >>In fact, an earlier version >>(5.0) defeated three Grandmasters in a single six round event (Aegon) at >>standard time controls. At fast chess, it is a terror;" (Kaufman Nov, 1998) > >Is Larry Kaufman the definitive word on whether computers are >Grandmaster strength yet? YES, almost. You don't know this guy >the way I do. Although everyone is fallible, I would have a hard >time pointing to a more reliable source than this to answer this >question. Larry knows computer chess like the back of his hand. >He knows the problems and weaknesses but he also is a strong >player although he is not a Grandmaster. His opinion on this will >be better than almost any grandmaster opinion. Why do I make this >ludicrous statement? Because it doesn't have as much to do with >strength as it does to objectivity and rational thinking. Larry >is king in this regard and knows ALL the facts and is not prone >to making hasty or biased judgements. I don't think he has a >biased bone in his body! He is simply a clear rational thinker >and is ultimately qualified to evaluate chess programs. > >So if Larry said this, I wouldn't bet against it, unless you just >want to throw your money away, in which case you can just send it >to me and save yourself the hassle! > > >- Don I'd never personally doubt Larry's remarks or intentions as having any sort of bias or personal agenda. However, he does have a weakness... in that he doesn't work with all sorts of computers all the time. IE we had some long conversations at the Indianapolis ACM event, and he simply didn't understand just what Cray Blitz was all about or what it was capable of... it was simply outside of his experience. When he tossed some test positions at it, it ate 'em up in zero seconds per position, and he wasn't quite expecting that... because he really didn't understand the difference between a Cray and a 486/66, except that the cray was big and expensive and the 486 was small and cheap. :) But he learns, and he asks questions. And he gets surprises as we all do... one simple case in point was the "pawn race" code in CB... so he tried the obvious trick of giving white an A pawn, black a H pawn, and letting them race. CB knew that which ever one moved first won (as does Crafty) because it sees that one queens and prevents the other from queening. We then moved the "loser's" king down to protect the queening square and voila.. CB said "draw" (as does Crafty). Then we tried other variations... one favorite was where the position seemed drawn, but the side queening first could then check the king and when it moves away rip the pawn that had just promoted... (Crafty doesn't do this)... we had a lot of fun trying to find a case where it would screw up with a 1 ply search... but it worked well. His main comment was "How can you afford to do that in your eval, since you can see all that with a 1 ply search?" My response was "the miracle of vector processing..." :) Bob
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.