Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strength of the engine in chess programs

Author: Robert Hyatt

Date: 15:20:59 05/26/02

Go up one level in this thread


On May 26, 2002 at 11:35:40, Rolf Tueschen wrote:

>On May 25, 2002 at 23:30:29, Robert Hyatt wrote:
>
>>On May 25, 2002 at 13:06:00, Rolf Tueschen wrote:
>>
>>>On May 24, 2002 at 11:42:21, Robert Hyatt wrote:
>>>
>>>>On May 24, 2002 at 08:31:59, Rolf Tueschen wrote:
>>>>
>>>>>Thanks.
>>>>>
>>>>>We have two open issues.
>>>>>
>>>>>1. My question about the first game in '97. Why Kasparov couldn't repeat his
>>>>>dominating play against DB2? Was a general question about the apparent twist in
>>>>>the match after the first game. BTW the mistakes of the machine were widely
>>>>>analyzed and "accepted", as far as game 1 is concerned. Could you comment on the
>>>>>mistakes, the twist and their reasons?
>>>>>
>>>>
>>>>
>>>>
>>>>
>>>>I still don't understand the question.  Kasparov won game one, in what I
>>>>consider to be a _very_ lucky manner.  He lost an exchange.  He just happened
>>>>to end up with a couple of passed pawns that were not stoppable over time.  I
>>>>am not sure he planned on that happening, but it did.  He didn't try to repeat
>>>>game one again.  In game two the colors were switched.  In game 3 he tried a
>>>>different opening.
>>>>
>>>>So I don't know what you mean by "why kasparov couldn't repeat his
>>>>dominating play against DB2?"  I don't see where he tried...
>>>
>>>I was just asking for your opinion about the world-wide accepted judgement that
>>>DB2 played mistakes over and over in game one. Consequence the domination by
>>>GKasparov. My question was not in regard to the opening but to the strategical
>>>play of Kasparov. But if you see no rational in the question, we could let it
>>>drop. I was just interested.
>>
>>
>>I don't see much proof it played "mistakes" over and over in game 1.  Moves
>>criticized by others were labeled as "the only reasonable move" by Kasparov.
>
>So, nothing new. That was my question if something new had been found. Also from
>the DB2 side. I have still other GM's opinion in mind. It's not just usenet
>blabla. I still don't understand the principal change (if ever) between game 1
>and 2. But ok.

According to the DB team members, no change was made between the first two
games (something was changed later in the match, but I don't remember
specifics).  And since changing between round 1 and round 2 was perfectly
acceptable practice, I don't see why they would say "no change" if they
really did change something...



>
>>
>>
>>
>>
>>>
>>>
>>>>
>>>>
>>>>>2. The Eduard Nemeth bet. Ok, fine. I talked to him via chat and he says he
>>>>>wants 30 minutes games. But since he hasn't really understood yet our debate
>>>>>here, he tought that it was a challenge from your side out of the blue, and his
>>>>>first question was "how much money for each point". Well, I promissed him 200 ?
>>>>>(euro) from _my_ side, but explained that 10 game only wouldn't be enough, it
>>>>>should be about 20 or 30 games overall. Could you please comment a little bit
>>>>>from your knowledge about similar "bets" on ICC and the probable outcome? I mean
>>>>>would you say that GM Roman would win 1 pt. pro 10 games trivially or is that
>>>>>already difficult for him in 30' games?
>>>>
>>>>
>>>>Roman would _definitely_ win one of every 10 games.  Against _any_ program.
>>>>And that would be a lower bound.  I would suspect he would win at least
>>>>4 and perhaps do better depending on how well his opponent was able to avoid
>>>>the kinds of positions Roman is good at setting up.
>>>
>>>Are you talking about 30' rapids? I didn't find too much other than Bullet and
>>>Blitz in you very good database.
>>
>>I don't count blitz.  It is not "solved" yet but computers are strong enough
>>at blitz to make the contest not really interesting.  I was thinking of 30
>>minute games with an increment to prevent sudden-death.
>
>Fine! That is what Eduard was saying too. He didn't mention increments but we
>will see. Actually he's going underground from chess to lyrics... (cough)

He can play 30 0 if he wants...


>
>
>
>>
>>
>>
>>>
>>>>
>>>>>
>>>>>Another technical problem was for Eduard, who never played on ICC, that he
>>>>>thinks he could only play on the new FRITZ server. Well, what is your opinion to
>>>>>the technical problems. Could it be made possible to connect the two servers for
>>>>>the little event? Or would you also say, what I have told Eduard, that the two
>>>>>servers are probably similar for the players. Ahm, I remember, Fritz server has
>>>>>a direct connection between the players, where you know exactly how long the
>>>>>delay and stuff like this. Please give us your judgement, it could be
>>>>>interesting for many potential players from Europe.
>>>>
>>>>
>>>>He can easily download a windows version of winboard, install it, and
>>>>connect to ICC instantly.  It is very easy to set up and run.  There are
>>>>other ICC clients as well...   zics, blitzen, you-name-it...
>>>>
>>>>He can log in as a guest to avoid joining, or he could do the free-week deal
>>>>with no cost...
>>>
>>>Yes. Please give me another cople of weeks for Eduard, who is in a personal
>>>vacation actually. It seems to be easier to intrumentalize GMRoman than our
>>>Eduard. I did it for him besides my scientifical interest, but he is thinking
>>>that you are the basic challenger for him. :)
>>
>>Remember that the "challenge" really wasn't my idea.  But I am willing to
>>try most anything for fun...
>
>Yes, I know, but not Eduard. It was _my_ dumb idea. But he isn't ready right
>now. Pavarotti also has his problems at times. :)
>
>
>>
>>
>>>
>>>
>>>
>>>>
>>>>>
>>>>>And to the chess results, would you mind giving a border line, Eduard in your
>>>>>opinion could not overstep, e.g. 2 pts. or something? Final question, and please
>>>>>just for the German player who gives so much heart blood into computerchess,
>>>>>what could be the prospects for someone like Eduard (understood that he might
>>>>>even win 3 pts/10 games) in the world of CC?
>>>>>
>>>>>Thanks a lot for your answers
>>>>>
>>>>>Rolf Tueschen
>>>>
>>>>
>>>>What is Edward's FIDE (or equivalent) rating?  If he is (say) 2150, then I
>>>>would be surprised if he won two games, or if he drew four.  Or any combination
>>>>to produce 2 points in 10 games.
>>>
>>>Yes, he has about 2150 on a national level. Perhaps it's also interesting you
>>>that Eduard did not present so many victories about Crafty than about Fritz. So
>>>it might well be that certain things are not that easy against your prodigy.
>>>
>>
>>
>>I have beaten Crafty from time to time.  If I published just those games, it
>>would paint a picture that is _highly_ misleading.  However, if I published
>>those games just to show where it had a miserable understanding, it would be
>>possible that someone would conclude that I won nearly every game by only look-
>>ing at the result of each game I posted.  I suspect that Nemeth was posting
>>games to show program problems, rather than to say "I win every game."  We will
>>see...
>
>This is again about my interest in the title above. BTW Eduard is not a con man.
>But IMO his discoveries reveiled many structural weaknesses of the computers.
>But I know that you are the first to agree here. But I don't know if you know
>how widely accepted the opinion is of the "2700" programs. Here you are almost
>an alien with your opinion of the only 2500. And for me >Eduard is a symbol for
>even deeper "strength", because traditionally we say that the strength is
>depending on the weakest member.  There are still too many weaknesses.

The ones claiming 2700+ simply don't understand 2700+ chess, so those are
just "numbers".   2500 is far closer to reality than 2700.  If 2500 is a
bit low, 2700 is _way_ high.  So long as we talk about traditional chess
and not 30 minute games.




>
>>
>>
>>
>>
>>>
>>>>
>>>>I don't think computers are "magic" in any way.  But a 2100-level player, while
>>>>able to win a game here and there, is going to be hard-pressed to win a single
>>>>game against a reasonable computer opponent.  I am, of course, curious whether
>>>>he can actually do so or not, from an academic point of view...
>>>
>>>For Eduard it's more a question of honour right now against your challenge.
>>
>>
>>Please explain this challenge to him.  I don't doubt he can give computers
>>great problems.  Perhaps greater than his 2150 rating would suggest.  But I
>>think he will have his hands full playing a series of games where every game
>>counts..
>
>More than this. His problem is the time of the training. But this is nothing for
>joking. This is what we said. With the necessary training things look different
>against computers. You never said different.
>
>
>>
>>
>>
>>
>>>Well, I'll keep you informed if I hear something new after the delay. Could you
>>>in return inform me about a special historical event at ICC, where some
>>>anononymous player "mercilious" did win some games in a row with a special trick
>>>and who was banned for that (?) reason by ICC? So, basically it's not supported
>>>if human players try such gambling? I have thus far read the news in Tim Krabbé.
>>
>>
>>This was the "trojan horse attack."  He got into trouble because he used it
>>against all the comps on ICC, beating them over and over in the same game,
>>simply to inflate his personal rating there.  He would play the same game
>>10 times if the computer would allow it.
>
>Was this the rule, was it forbidden to do? That was my astonishment.Was it
>really a problem for you after the learning files (Uri made the point already)?
>
>Rolf Tueschen

It is called "rating abuse" and takes many forms.  One player intentionally
losing to another to raise his rating.  Doing the opposite against computers
was another specific no-no covered in "rating abuse" on the chess servers.




>
>>
>>
>>
>>
>>
>>>
>>>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.