Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Typical delusions in CC

Author: Rolf Tueschen

Date: 13:02:14 01/26/03

Go up one level in this thread


On January 26, 2003 at 15:52:32, Uri Blass wrote:

>On January 26, 2003 at 15:29:30, Rolf Tueschen wrote:
>
>>On January 26, 2003 at 14:21:26, Mark Young wrote:
>>
>>>On January 26, 2003 at 14:06:05, Rolf Tueschen wrote:
>>>
>>>>On January 26, 2003 at 13:49:14, Mark Young wrote:
>>>>
>>>>>On January 26, 2003 at 10:29:26, Rolf Tueschen wrote:
>>>>>
>>>>>>On January 26, 2003 at 06:27:51, Frank Phillips wrote:
>>>>>>
>>>>>>>On January 26, 2003 at 05:30:01, Roger D Davis wrote:
>>>>>>>
>>>>>>>>>Unless I have misunderstood the contract, this is a modified version of chess to
>>>>>>>>>that defined by the standard rules, which nowhere states that because one side
>>>>>>>>>knows that the game is drawn with perfect play then it shall be declared a draw
>>>>>>>>>- even if the other side does not know or cannot demonstrate it.
>>>>>>>>
>>>>>>>>Presumably then, Kasparov could show up with his own set of tablebases, and
>>>>>>>>consult them during the match? Maybe he has a particular ending he's weak in. Or
>>>>>>>>do only Junior's tablebases count?
>>>>>>>
>>>>>>>I thought this was man versus machine.  Humankind advantages and disadvantages
>>>>>>>versus machine advantages and disadvantages.  See who prevails.
>>>>>>
>>>>>>This is a good example for delusions and pure wishful thinking in CC. At first
>>>>>>this sounds absolutely in order. I will prove now why it is in reality a cheat
>>>>>>with a logical fallacy.
>>>>>>
>>>>>>Let's describe the conditions.
>>>>>>
>>>>>>I. We have a long tradition in human chess. Rules have been made. These rules,
>>>>>>normaly that must not be pointed out, are made for two human chessplayers. In
>>>>>>special to prevent that one player takes unfair advantages from outer help.
>>>>>>Books, other documents or conversation with collegues.
>>>>>>
>>>>>>II. In computerchess people saw directly that without "books" taken from human
>>>>>>chess computers could not play sound chess at all. I mean alone based on their
>>>>>>engine [which is different nowadays, at least a bit]. So it was clear, also in
>>>>>>the understanding of computer sciences, that files, yes, whole databases could
>>>>>>well be added to the chess engine. Today endgame tablebases have been
>>>>>>successfully implemented so that the computer is now prepared to play perfect
>>>>>>chess the moment he gets the access to the tables. This is already possible long
>>>>>>before the concrete chess position is on the board. Consequence: chess is being
>>>>>>played only in between the zones of perfect knowledge [ok, not for me because I
>>>>>>often discussed that for the super GM the given opening theory is always the
>>>>>>picture from the past, but not the actual possible because this is exactly
>>>>>>researched by these masters - and then quickly copied and pasted by the CC
>>>>>>people; but it is clear thaqt the GM have always the advance].
>>>>>>Now with a certain irony CC people sigh that in the past no GM ever complained
>>>>>>that something is wrong with the addings, but the moment they began to lose,
>>>>>>they were against these hybrides. Is this a correct picture of the real
>>>>>>situation? Of course not.
>>>>>>
>>>>>>III. The truth is that
>>>>>>a) for comp - comp matches the addings are ok in a way [but only in a way
>>>>>>because that would be nonsense if the engines would be equally strong and the
>>>>>>differences would be defined only by the addings, books and learning features
>>>>>>etc.]
>>>>>>b) for comp - human matches we should ask which rules are respected. Strictly
>>>>>>after the FIDE rules books and tables would be forbidden.
>>>>>>
>>>>>>Let me explain why the situation in b) is extremely unfair for the masses of
>>>>>>normal players in both respects (book and tables) and even for super GM in
>>>>>>respect of the tables, always because of the perfect play while humans, even
>>>>>>super GM are unable to play perfectly [depending on the difficulty of endings].
>>>>>>
>>>>>>IV. We have a logical fallacy if we simply state that a match between comp and
>>>>>>human should be played so that each side plays after the rules of its side.
>>>>>>Because FIDE rules are made for humans only while computer rules are
>>>>>>traditionally made for both,  pure comp and also comp - human matches. I thi
>>>>>>nk it's clear that this handling is unfair. We must find rules for comp - human
>>>>>>matches. And for the Kasparov - Junior match a solution has been found. It
>>>>>>should rule out that the human player can lose an objectively even (=drawn)
>>>>>>game. I read that people in CC said that this would be ridiculous because you
>>>>>>can well lose a drawn ending as a human. I say that this is correct but this
>>>>>>would give the comp side an unfair edge. It's a question of memory, because the
>>>>>>chess engine does NOT calculate the moment it has found the draw but it can look
>>>>>>into the tables. This would be the same if the human GM had all these tables
>>>>>>moves in his memory which is impossible. But by force we must also conclude the
>>>>>>same for the amateur players who play with a very limited memory in the opening.
>>>>>>So we should change the rules so that also amateur players can use books. BTW
>>>>>>most players do this already when they play against computer programs at home.
>>>>>>If the comp sie would argue that this would be against the rules, then humans
>>>>>>simply answer that this is not about human chess but human - comp chess and
>>>>>>there the comps are allowed to use books. To say that this is ok, that comps are
>>>>>>allowed but that humans are NOT allowed is again a logical fallacy. Because the
>>>>>>CC people then would take advantages out of a sphere that is not their own. But
>>>>>>what I do at home is my stuff and none outside can interfere. But if we meet
>>>>>>then we must find rules for our meeting. And it would be nonsense to follow the
>>>>>>guide that humans must then play as it is in human chess. I think this should be
>>>>>>clear by now. Thank you for your attention.
>>>>>>
>>>>>>V. Let's give a perspective for the future. Since the zone where real chess is
>>>>>>being played is so small we should change the rules still further. We should
>>>>>>either forbid books for comps or should allow books for humans too. The latter
>>>>>>should be the easier in practice. Because you can't control that the engine has
>>>>>>no implementations regarding books. - But all these reflections are moot because
>>>>>>a concrete tournament play with comps is only fantasy. The main reason is [and
>>>>>>also this has been discussed in many articles, also by me] that in longer
>>>>>>periods human players would adapt on the comps play and very quickly comps would
>>>>>>be shown as what they really are, namely very stupid machines. Simply because no
>>>>>>matter how deep they could calculate humans can adapt to certain weeaknesses and
>>>>>>find typical weapons to exploit these weaknesses. Since all these have something
>>>>>>to do with depth, hence the consequences will be fatal until chess will be
>>>>>>solved in the year 5000. So by force the only play will be in such show events
>>>>>>where the human GM gets the neccessary money for a commercial interest, namely
>>>>>>the influencing of users who might be cheated about the real strength of the
>>>>>>product. Since such events [like Kasparov vs Junior] allow many players to get
>>>>>>some money as commentators or whatever, the truth about the real situation,
>>>>>>regarding strength, will never be told by named experts. Those who know the
>>>>>>truth and could say the truth are not taking for serious in the world of fantasy
>>>>>>and wishful thing, not to forget in a world where money is so important.
>>>>>>
>>>>>>I wish us, me included, all a good Sunday of CC.
>>>>>>
>>>>>>Regards,
>>>>>>Rolf Tueschen
>>>>>
>>>>>It is not unfair for the computer to use its own programmed memory in the
>>>>>opening and endgame. This is just the nature of the beast when you play a
>>>>>computer. Many humans who happen to be strong GM's also use book info and commit
>>>>>it to memory. I don't know of any good player that has not used opening
>>>>>information based on other players opening works to help his own game.
>>>>>
>>>>>Chess is far from played out. If the human wishes he can go into perfectly
>>>>>playable lines far from move 10 to make the program start thinking in the
>>>>>opening phase of the game.
>>>>
>>>>
>>>>Please do not leave out the connectivity of my statements. Taken isolated
>>>>nothing is unfair in CC, but I spoke of competition with human players.
>>>>
>>>>Rolf Tueschen
>>>
>>>I also spoke of competition with humans. You can argue anything, but to state
>>>that computers do something that the humans players don't do is clearly wrong.
>>>The computers are just perfect at known theory and simple endgames,
>>
>>Let's keep in mind that we are just discussing here. That means if we see errors
>>in the other's arguments we do NOT say that the other is a fool or has no
>>knowledge. -
>>
>>But now let me state that this is wrong when you write that "computers are
>>perfect in known theory and simple endgames".
>>
>>Two reasons:
>>
>>1) How can you assume perfection IF theory contains many errors?
>
>I agree that computers may play errors when they are in book but it does not
>mean that it is unfair to let them to take the risk of playing the errors.
>
>
>>
>>2) How can you say that, if the engines (without book) would be unable to play
>>correct openings, at least in so called difficult opening with sacs in special?
>>As I said earlier, isn't it a cheat to pretend something you can't do in
>>reality? Isn't it impostering?
>
>cheating is doing something that is not allowed by the rules.
>playing moves that you do not find by search because they are part of the book
>is allowed.
>
>The computer does not pretend to search when it looks in a database.


You did misinterprete me, I said that the pretension is 'that the computers play
with book to keep it simple and fast but in real they also could find the same
moves on their own'. I say that this is a cheat because they can't do that in
still _many_ cases. I know that you would not object here. :)

Rolf Tueschen



>
>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.