Author: Robert Hyatt
Date: 14:28:44 07/08/01
Go up one level in this thread
On July 08, 2001 at 14:13:24, Uri Blass wrote: >On July 08, 2001 at 10:20:28, Robert Hyatt wrote: > >>On July 08, 2001 at 06:31:45, Uri Blass wrote: >> >>>On July 08, 2001 at 00:13:06, Robert Hyatt wrote: >>> >>>>On July 07, 2001 at 23:21:52, odell hall wrote: >>>> >>>>>> >>>>> I really don't understand what your saying, since even on icc computers are >>>>>performing like supergrandmasters. >>>> >>>> >>>>I'm sure you don't. But when you have time, point out a single "super GM" >>>>that loses to a 2000 player. The computers are good, no doubt. But they >>>>are not _that_ good yet... >>> >>>1)I do not know about super GM's who lost against 2000 players but I am going to >>>be surprised if there were no cases when it happened. >> >>What about the case where a computer loses reasonably often? And draws 1/3 to >>1/2 of the games? A GM will certainly make an occasional blunder, or overlook >>some deep tactical shot that the 2000 player will on a few occasions find. But >>if a 2000 player simply outplays the machine game after game, only to make a >>tactical mistake here and there, then that computer is going to have a hard time >>convincing me it is a GM, even if it wins _every_ single game against that >>player. >> >>A 2000 player is not going to positionally outplay a GM game after GM only to >>lose after making a tactical error. >> >> >> >> >>> >>>I know about 2 cases in Israel when GM's lost against 2000 and 2100 players at >>>tournament time control and I am sure there are more that I do not know. >>> >>>In the case of the loss against the 2100 I know that everything was opening >>>preperation. >> >>I don't even count that kind of thing... Against the computers on ICC I am >>talking about a different thing altogether. >> >> >> >>> >>>In the case of the loss against the 2000 player it was not opening preperation >>>and the GM simply did not play well. >>> >>>I guess that super GM's play less games against 2000 players relative to >>>computers and this is one of the reasons that they lose less games. >>> >>> >>>2)It is known that computers have weaknesses but it proves nothing. >>> >>>If a computer program scores 99% against 2000 players and 85% against GM's then >>>I think that it is right to say that it is performing like a super GM and the >>>fact that the super GM may get 99.9% against 2000 players does not change it >>>because the super GM may get less against GM's. >>> >>>Uri >> >> >> >>If you look at only results, (mainly blitz games on ICC) to predict how longer >>time controls will go, that is a mistake. A human will totally outplay a >>computer at blitz on ICC only to lose to a tactic he didn't have time to >>examine fully. As the time controls stretch out, the human's positional skills >>don't particularly get better (neither does the computer's) but the tactical >>errors are reduced. > > >The positional mistakes of the computer are also reduced when the time control >is longer. > >When programs search deeper they can find better positional moves because they >can see by search positional advantage that they do undersatnd. Only in some cases. IE you allow your opponent to create a queenside majority with abc pawns to your bc pawns. In doing this, you win a pawn on the kingside. If you don't know the potential danger of that candidate passer on the queenside, you are going to make serious judgement mistakes. And if you have to search deep enough to actually see the passer on the board (or even worse, to see the promotion event itself) then the computer is going to fall on its own sword. Because the depth required is simply too great for the next 100 years. > >For example Junior likes to move out with the queen early >when it is out of book at blitz but it can play better at longer time control > >Example: >After 1.e4 d5 exd5 Qxd5 a3 Junior7 likes to play Qe4+ that is not >a good positional move because of Ne2 and Deeper search helps it to develop >pieces That is simply a loss of tempo issue. I am thinking about much more serious issues with pawn structure and king safety. And recognizing an endgame that is unlosable, or unwinnable, or better. > >Blass,U - Junior 7 >rnb1kbnr/ppp1pppp/8/3q4/8/P7/1PPP1PPP/RNBQKBNR b KQkq - 0 1 > >Analysis by Junior 7: > >3...h5 4.Qf3 > = (-0.19) Depth: 3 00:00:00 >3...h6 4.Qf3 > = (-0.25) Depth: 3 00:00:00 >3...Nh6 4.Qf3 > ³ (-0.32) Depth: 3 00:00:00 >3...Nf6 4.Qf3 > ³ (-0.48) Depth: 3 00:00:00 >3...Nf6 4.Nc3 Qe6+ 5.Qe2 Nc6 6.Qxe6 > = (-0.22) Depth: 6 00:00:00 2kN >3...Nc6 4.Nc3 Qe6+ 5.Qe2 Ne5 > = (-0.25) Depth: 6 00:00:00 2kN >3...Qe4+ 4.Qe2 Qxc2 5.Nc3 Bd7 6.Qe5 > ³ (-0.33) Depth: 6 00:00:00 11kN >3...Qe4+ 4.Qe2 Qxc2 5.Nc3 Nf6 6.Qd3 Qxd3 7.Bxd3 Nc6 > ³ (-0.27) Depth: 9 00:00:00 42kN >3...Qe4+ 4.Qe2 Qxc2 5.Nc3 Nc6 6.Nf3 Bg4 7.h3 Be6 8.Qd3 > = (-0.19) Depth: 12 00:00:01 812kN >3...Qe4+ 4.Ne2 Bf5 5.d3 Qe6 6.Be3 Nc6 7.Nf4 Qe5 8.d4 Qa5+ 9.Bd2 Qb6 10.Bc3 > = (-0.06) Depth: 15 00:00:39 22073kN >3...Nc6 4.Nc3 Qe6+ 5.Qe2 Nd4 6.Qxe6 Bxe6 7.Bd3 0-0-0 8.Nge2 Nb3 9.Rb1 Nf6 > = (-0.07) Depth: 15 00:00:44 24924kN >3...Bf5 4.Nc3 Qe6+ 5.Qe2 Qxe2+ 6.Ngxe2 > = (-0.10) Depth: 15 00:01:31 52629kN >3...Bf5 4.Nc3 Qe6+ 5.Qe2 Qxe2+ 6.Ngxe2 Bxc2 7.Nd4 Bg6 8.Nd5 Kd8 9.Bc4 Nd7 10.h3 > = (-0.13) Depth: 16 00:02:01 68599kN > >(Blass, Tel-aviv 08.07.2001) > >Uri
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.