Author: Rolf Tueschen
Date: 12:25:46 02/04/03
Go up one level in this thread
On February 04, 2003 at 14:11:09, Chris Carson wrote: >On February 04, 2003 at 13:31:00, Dann Corbit wrote: > >>On February 04, 2003 at 13:10:57, Chris Carson wrote: >> >>>On February 04, 2003 at 09:12:12, Rolf Tueschen wrote: >>> >>>>On February 04, 2003 at 08:21:26, Uri Blass wrote: >>>> >>>>>On February 04, 2003 at 06:40:16, Rolf Tueschen wrote: >>>>> >>>>>>On February 04, 2003 at 05:51:17, Andreas Guettinger wrote: >>>>>> >>>>>>>On February 03, 2003 at 19:05:27, Rolf Tueschen wrote: >>>>>>> >>>>>>>>On February 03, 2003 at 18:54:54, Peter Hegger wrote: >>>>>>>> >>>>>>>>> >>>>>>>>>...how is it that they now consistently play at the 2700-2800 level? Against >>>>>>>>>Kramnik (2810), against Bareev (2729), and now against Kasparov (2807), a >>>>>>>>>program is turning in a 2807 performance and very much _holding its own_ >>>>>>>>>Calling any modern program a 2500 player is akin to calling the above mentioned >>>>>>>>>super GM's 2500 players. >>>>>>>>>It also looks to me as though the SSDF list is getting closer to the reality of >>>>>>>>>the true state of program prowess than (admittedly) it use to be. >>>>>>>>>Any comments welcome. >>>>>>>>>Regards, >>>>>>>>>Peter >>>>>>>> >>>>>>>>A pity that you do not read. Show events are NOT a possible tool to calculate >>>>>>>>the strength. And hard competition doesn't exist. That's it. I still hold >>>>>>>>that comps are 2400 at best in fierce tournament chess. >>>>>>>> >>>>>>>> >>>>>>>>Rolf Tueschen >>>>>>> >>>>>>> >>>>>>>First of all, I agree that humans play a lot better chess than computers. I.e. >>>>>>>DJ did not mananage to get a reasonable attack against Kasparov in the current >>>>>>>match. Humans seem to be much mure creative chess players. >>>>>>> >>>>>>>On the other hand, I doubt that the average rating performance of computers are >>>>>>>much less than that of humans. Or say, they have the same Elo than top players. >>>>>>>I.E. the human plays a top game with 47 good moves on a high level and 3 >>>>>>>blundering moves in it and will lose against the computer. >>>>>> >>>>>>I think you make a couple of typical mistakes in your chain of reasoning. You >>>>>>completely leave out the idea of strict anti-comp play and I am talking about >>>>>>strict = played and tested over a longer period of time. By many, not just one >>>>>>or two players. Also you misunderstand Elo. Why should Elo of the players go >>>>>>down when they control comps? My 2400 for comps (2003!) is a "vision" if you >>>>>>want IF certain conditions would exist. And now let's take a look at chess. >>>>>> >>>>>>You say that a human makes 3 weak moves per game. I say you are talking about >>>>>>human chess with all its delusions and projects. Here however we are talking >>>>>>about anti-comp. Now think for a moment. You say that comps play not bad but >>>>>>without great highlights. I think we all know that from the imbreeding campaigns >>>>>>in SSDF etc. What does that mean? Very simple. The human anti-comp can just >>>>>>concentrate on the typical anti strategies. There is no hurry. There is no >>>>>>tension. There is no creativity on the side of the machines. What - honestly - >>>>>>will you have more??? >>>>>> >>>>>>I think that most people underestimate the human brain. And surely that of GM. >>>>>>Look. A GM has a computer in his brain! How many times must I repeat that? You >>>>>>become GM because you have that inborn talent of outstanding memory capacities >>>>>>-first. And that is a must! Sine qua non - for all scientists. And then of >>>>>>course the necessary talent for chess, which should be discovered with the age >>>>>>of 5 or 6 years. Now the irritating news from psychology is: that memory has not >>>>>>too much to do with say artistic genius in general. It could well be that you >>>>>>land in a boring bureaucratic job but you know all the numbers of the telephone >>>>>>book of say NY. That is also why chess GM are not by force deep thinkers. Not to >>>>>>speak of philosophers. Lasker is the exception. >>>>>> >>>>>>That aspect is important because I must explain here why up to now not too many >>>>>>GM really started to perform against comps. And those few who perform, do that >>>>>>in show events with the usual hoax we know from simuls and other exhibitions. >>>>>>That is mostly about money nothing else. My theory is that IF a few clever GM >>>>>>would begin to compete against comps we would realise very fast how weak the >>>>>>machines are in reality. IF humans develop a special 'counter technique'. But >>>>>>make no mistake: you must be able to calculate lines up to say five moves. So >>>>>>all amateur players and patzers below 1700 bye bye. :) >>>>>> >>>>>>Know what I mean? Actually we have a complete fog of hot air in CC because NEVER >>>>>>at least in public that has been shown by GM how serious the weaknesses of comps >>>>>>really are. In public and for money it's part of the deal that GM simply don't >>>>>>touch it what is weak. Or don't _talk_ about it. >>>>>> >>>>>>So, perhaps now you know when the number 2400 could exist and in which >>>>>>conditions. >>>>>> >>>>>> >>>>>>>If it is a 2800 >>>>>>>player, then the computer may have played the weaker moves, but won in the end >>>>>>>and deserves reating of that level too. >>>>>> >>>>>>Fine. But the comps would also lose against 1900 or 2100 player! And that would >>>>>>NEVER happen to GM! Period. :) >>>>> >>>>>losing against 1900 or 2100 player can happen to a GM and I know of cases when >>>>>it happened. >>>>> >>>>>In one of them the GM fell to a prepared trap that the 2100 discovered in >>>>>analysis before the game. >>>>> >>>>>In another case the GM simply did not play well and did mistakes so he lost >>>>>against 1900 player. >>>>> >>>>>Uri >>>> >>>>Please give exact headers or scores. And we were talking about 2600-2700 GM not >>>>2445 GM from institutions for seniors. Please also exclude first rounds in Opens >>>>because that is show event and commercial. :) >>>> >>>>Rolf Tueschen >>> >>>This Just ended 1/31/03: >>> >>>Auckland International Open Chess Tournament - Pairings and Results >>> Round 9 Results >>>No Name Fed Rtg Result Name Feder Rtg >>> >>>1 McLaren, Leonard NZL 2274 1:0 GM Sermek, Drazen SLO 2577 >>>2 FM Watson, Bruce NZL 2266 .5:.5 IM Froehlich, Peter GER 2421 >>> >>>There are a lot of these, just look in TWIC, you will find 300 point underdogs >>>winning or drawing games all the time, this is just the latest example. There >>>are also plenty of examples of larger ELO differences with the lower rater >>>player winning. McLaren has no title, but does have the win over a GM. >> >>The problem with finding 2600+ players against sub 2000 players is that they >>don't meet very often at 40/2 OTB. >> >>With a 600 Elo difference, only 3% of the points will be won by the weaker >>player, and it will be hard to find hundreds of such games to make finding one >>likely. >> >>How many serious tournament games does Kasparov have against sub-2000 players in >>the past 10 years? I would guess not very many. > >Yes, 2600+ vs 2200- is very rare, even for Open events. However, a 1500 player >can play Fritz7 on a 2Ghz machine as often as they like. Find weakness, learn >good anti-computer skills and get a score that is higher than predicted (if for >no other reason than resetting the program and starting over with the new >knowledge). > >I have doubts that a below 2400 player has much of a chance against DJ, DF, >Hiarcs, Shredder or CTiger on 3Ghz machines or faster under these match >conditions (let alone tournament format) with the programmers running the >program. > >Best Regards, >Chris Carson Could you please elaborate why anti-comp shouldn't work for Eduard-like people? Rolf Tueschen
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.