Author: Rolf Tueschen
Date: 05:51:16 02/04/03
Go up one level in this thread
On February 03, 2003 at 20:22:27, Uri Blass wrote: >On February 03, 2003 at 17:31:36, Rolf Tueschen wrote: > >>On February 03, 2003 at 08:12:50, Uri Blass wrote: >> >>>On February 03, 2003 at 07:56:27, Rolf Tueschen wrote: >>> >>>>On February 03, 2003 at 07:44:27, Uri Blass wrote: >>>> >>>>>On February 03, 2003 at 06:54:19, Rolf Tueschen wrote: >>>>><snipped> >>>>>> We all >>>>>>know that a chain is only as strong as its weakest member. >>>>> >>>>>No >>>>> >>>>>I do not know it. >>>> >>>> >>>>But this is not my fault then, Uri. It's simply knowledge. Very old knowledge. >>>>If you have a giant to fight, then you look, where he's vulnerable, no??? And IF >>>>his vulnerable spot or habit is constant, fixed, you start to treat exactly that >>>>spot. You are not stupid and start on his strong, too strong, sides. A chain >>>>will always break in its weakest member. Try it. Open a bit one member, so that >>>>it's weak compared with the others, then you will see the result. >>>> >>>> >>>> >>>> >>>>> >>>>> Who could seriously >>>>>>doubt that? But in computerchess that happens every day! People create a huge >>>>>>confusion just to tweak the truth. And the truth is that because chess programs >>>>>>have weaknesses even a 1700 player wouldn't have, they can't be - by definition >>>>>>- Grandmaster or beware Super Grandmaster, they are not even International >>>>>>Master - - - _IF_, and only if the mentioned groups would be in a serious >>>>>>competition with such an _individual_ chess program and NOT always with a little >>>>>>b >>>>> >>>>>By your theory computers cannot be even 1700 players but it is obvious that even >>>>>better players who tried to play against them find that they have no chance >>>>>because they canot get the positions that they understand better in the board or >>>>>cannot aboid losing in position when they had the advantage. >>>>> >>>>>Without discussing the question about the exact level of computers I think that >>>>>nobody will claim that computers are weaker than 1700 players. >>>>> >>>>>Uri >>>> >>>> >>>>Unfortunately you snipped the major part of my message. So by definition you can >>>>deal with the isolated snips and get false perceptions. >>>> >>>>Chess is not a one-dimensional action. So if 1700 is the weakness, that weakness >>>>can't be explotated in 100% of the cases. So by force the Elo number will >>>>increase. Because we were in consense that comps could well be 2700 tactically. >>> >>>I do not think that there is a consensus that comp are 2700 elo in tactics. >>>I think that they are better than it in tactics. >>> >>>They never do tactical mistakes that 2700 players do also in human-human games. >>> >>>>So then you get my "estimated" 2400 in brutal competition. NOT in freaky show >>>>events of course. >>>> >>>>Rolf Tueschen >>> >>>If you get from 2700 and 1700 average of 2400 then it means that the weakest >>>chain is not dominant because 2400 is close to 2700 and not to 1700. >> >>I didn't calculate the average. All data just guesses. >> >> >> >>> >>>My opinion is that it is not 2700 and 1700. >>> >>>In tactical positions computers perform like 3000 players. >> >> >>Just guesses. Symbolic. Ok, I know what you mean. >> >> >> >> >>> >>>In most quiet position computer still perform like 2600-2800 players. >> >>You believa that? Only because Anand made a little bit PR for ChessBase you >>believe that? > >No > >It has nothing to do with anand. >The point is that search help not only in tactical positions. > >programs have some positional understanding in their evaluation and if you >combine the positional understanding with deep search you may get good moves in >most quiet positions. > >> >> >> >>>In minority of the positions they perform like 1700 players or worse but it is >>>hard to get them in practical games. >> >> >>How do you know? Eduard Nemeth knows some dozens or more variations to get a >>program into helplessness, of course 3000 tactical ON! But still the progs lose >>like 1700. Could you explain it? Let me help. Real masters simplx don't like >>such play. Ergo, they don't train. And that gives you the impression you >>described. > >Nemeth lose more games than he wins and analysis of his wins show that usually >the programs do not do the same mistakes at long time control on fast hardware. > >I am sure that GM's can beat programs by wrong sacrifices if they want to do it >but they will lose more games by this tries before they get their win. > >If they did not plan all the game before the tournament this is a big risk >to play wrong moves and it is better for hem to play for a draw against the >computer and not to try to win with probability of 10% by Nemeth's way when >there is 90% probability for a loss. > >> >> >> >> >>> >>>computers already performed better than 2400 in the israeli league when the >>>teams had the right to choose the human to play against the computer so they >>>could choose humans who are relatively better against computers. >> >>So you accept that training helps. Good. > >I believe that training can help humans. >I also believe that it is possible to teach humans to play better. >> >> >> >>> >>>Note that the software and hardware in the Israeli league was clearly worse than >>>the software and hardware that is used against kasparov so I guess that Junior >>>is clearly better than 2400 and I believe that it has at least 2700 level. >> >> >>And what do you say to those who claim that the speed is not a 1 to 1 effect on >>strength? > >I think that speed help(I do not know how much but it helps) but the point here >is also better software relative to the software in the Israeli league. > >> >> >> >>> >>>Fritz had the bigger problems in the Israeli league because it was a more >>>popular program so players prepared better against it(I know about one 2300 >>>player who lost against Rebel and drew with Fritz and when I talked with him he >>>admitted that he prepared against Fritz(something that he did not do against >>>Rebel) >>> >>>Note that having the program give humans unfair advantage because they can play >>>with the opponent when the opponent does not know about the games. >> >>Notable argument! >> >> >>> >>>I think that a fair match should allow the human to see games of the program but >>>not to play with it before the match. >> >> >>Yes Uri, you say that because you can't imagine what GM could learn out of >>games. But your forbiding of own playing is impossible. Look, what do you want? >>Serious competition? Or do you want only the alibi to then claim ridiculous >>strength? You must solve that problem on your own programmer side, you can't >>create forbidden zones to stay a virgin. >> >>But Israel must be paradise, Uri, I wished I could visit these events. Just >>average competition without the show character. Fabulous. Could I visit Israel >>without fear to be killed in the war? > >I live in Israel and I do not feel afraid of being killed. >I cannot be sure that nothing will happen but statistics tell me that accidents >between cars kill more people than the terrorists. > >I know that statistics tells me only about the past and not about the future >but if you do not assume that things will be different in the near future you >should be afraid more of being killed by other reasons and not by terror even in >Israel. > >> >>Finally: Are you Mendi Bael? Yesterday I asked you on playchess com. You can >>reply via email if you prefer. > >I do not know who is Mendi Bael or what is the meaning of the question. > >Uri Mendi is a person from Israel on the Fritz server in Germany who often is the chief of transmitted games. Then he gives nice analyses also with his progs I think and since you once wrote that you also operated in the championships in correspondance I wanted to know that. Mensi is ofter very humourous. It was just a question to break through the internet anonymity. Keep on doing your good work and your writing here, I read all of your messages. Also I observe how you become more and more a scientist with Movei testings. 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.