Author: Vincent Diepeveen
Date: 18:58:37 08/04/01
Go up one level in this thread
On August 04, 2001 at 11:31:56, Uri Blass wrote:
>On August 04, 2001 at 00:02:14, Vincent Diepeveen wrote:
>
>>On August 03, 2001 at 19:25:31, Eduard Nemeth wrote:
>>
>>>On August 03, 2001 at 11:19:07, Ed Schröder wrote:
>>>
>>>>Received from David Levy.
>>>>
>>>>Ed
>>>>
>>>>===================================
>>>>
>>>>The following will shortly be posted on the ICCA site.
>>>>
>>>>
>>>>
>>>>
>>>>Since announcing the above decision there have been some complaints from
>>>>programmers who had planned to enter with multi-processor machines. Different
>>>>views were expressed to me, including the suggestion that programmers be allowed
>>>>to enter both the single processor and multi-processor sections, and a
>>>>suggestion that we leave things as they were originally and have only one title
>>>>available no matter whether it is won on a single or multi-processor computer.
>>>>
>>>>In a situation like this no matter what is decided there will be complaints.
>>>>That is the nature of such situations.
>>>>
>>>>I have reflected on the various arguments made to me over the past couple of
>>>>days and on the wisdom or otherwise of making a further change in the rules, but
>>>>in the end I have decided that my original decision should stand, and this is
>>>>final. Although some of the multi-processor programmers will not be happy with
>>>>the decision I feel that it is the best one for a big majority of programmers.
>>>>And the multi-processor guys can still win a World Championship title.
>>>>
>>>>So please, no more complaints. Get your programs into a good mood for the event.
>>>>And good luck to all of you in Maastricht.
>>>>
>>>>I should emphasize again that my decision has been taken in the absence of the
>>>>Maastricht oirganising team so please do not blame any of them. Whatever fault
>>>>there is for this decision is mine.
>>>>
>>>>David Levy
>>>>
>>>
>>>My opinion "now"( german ):
>>>
>>>Nachdem die ICCA erst vor wenigen Wochen auch Multiprozessor Systeme zugelassen
>>>hat (was für eine Micro-WM nicht üblich ist, bisher wurden nur Systeme mit nur
>>>"einem" Prozessor zugelassen), hatte das Team Rebel um Programmierer Ed Schröder
>>>seine Teilnahme an dieser WM zurückgenommen.
>>>
>>>Vor wenigen Tagen wendete sich aber das Blatt, als der ICCA Präsident in einer
>>>Mitteilung die Regeln neu definierte.
>>>
>>>So wird es nun zwei Spielkategorien geben:
>>>eine für Multiprozessoren, und eine für Singleprozessoren! Demnach werden wir
>>>also auf jeden Fall 2 Weltmeister haben, plus einen Amateurtitel!
>>>
>>>Auf den ersten Blick erscheint diese Lösung für alle Beteiligten gut, aber nur
>>>auf den ersten Blick!
>>>
>>>Denn, würde ein Programm auf einem Singleprozessor mehr Punkte einfahren als die
>>>vermutlich auf "Dualsystemen" spielenden Programme, so wäre der Titel für den
>>>Multiprozessor-Titel kaum der Rede wert!
>>>Wird es anders ausgehen, so haben die "Singles" ein "Alibi" für die schlechtere
>>>Platzierung.
>>>
>>>Das Gelbe vom Ei ist diese neue Regelung für mich daher NICHT!
>>>
>>>Ich hätte es begrüßt wenn daher ALLE auf einem "Single" spielen müßten!
>>>
>>>Ich meine dass die ICCA einen Fehler begangen hat. Jeder Computerschach
>>>Interessierte konnte sich im letzten halben Jahr davon überzeugen dass die neuen
>>>Intelprozessoren (P4) für Computerschach nichts taugen.
>>>Für jene Programmierer die einen INTEL Prozessor favorisieren, bleibt der Dual
>>>die einzige Wahl um konkurrieren zu können. Die ICCA hätte daher schon längst
>>>die Duals zulassen sollen (und nicht erst vor wenigen Wochen), und keiner hätte
>>>dann etwas zum Meckern gehabt. Nun hat man also aus der Not eine Tugend gemacht
>>>indem zwei Titel vergeben werden, nur: gewinnen können so letztlich nur die
>>>Singles. Für Multiprozessor Systeme steht weit mehr auf dem Spiel!
>>
>>I think the opposite is true.
>>
>>Let's first list entries which are possible dual:
>> Crafty
>> Diep
>> Ferret
>> Fritz
>> Junior
>> Shredder
>> SOS
>>
>>Possible entering single title:
>> Gandalf
>> Gromitchess
>> Pharaon
>> Rebel
>> Spidergirl
>> Tao
>> Tiger
>> Xinix
>>
>>Now even if none of the single cpu machines say they want to qualify
>>for the multiprocessor title, then still there is a clear strength
>>difference in favour of duals.
>>
>>Chance that a single cpu machine is going to win this champ is real
>>small. The only possibility of this happening is if Shredder enters
>>the single cpu competition.
>>
>>Even then it will have a problem with 9 rounds to go.
>>
>>When we closely examine the single cpu list then we see a lot of
>>strong programs (only strong programs in world champ usually),
>>one of them, Rebel, could better next year ask to play for a senior
>>title, as now it's not going to win any title, but it might of course
>>help tiger by beating other single cpu stuff.
>>
>>This help must NOT be underestimated. Also with some major bookluck
>>it might win itself that single cpu title.
>>
>>Anyway times will be tough to let a single cpu win the tournament.
>>I estimate it at near 0% actually.
>>
>>And i'll explain why i think that's so low. First of all the only
>>real interesting program to get a high score in the single cpu list
>>would be tiger. This is the only program in that list which is
>>equipped with a real killerbook.
>
>How do you know?
>What about Gandalf?
>
>I also do not believe in killer books against experienced programmers.
>people can learn from experience.
>
>It is easy to get programs out of book in a few moves with equality with white
>or sligthly worse position but not losing position with black so if you are
>afraid from killer books it may be better to use small book against tiger with
>some not popular moves.
>
>For example you can start with 1.c3 with white and continue to build a book
>manually or to start with some lines like 1.e4 c6 2.d4 d6 with black and to
>continue to build the book manually.
>
>Uri
That's said by someone who didn't make a book for a program himself.
Someone who made books for a program would never say this.
Programs are complete idiots in opening. 1.c3 books are of course
complete nonsense. it's caro-kann with a tempo more if you're lucky,
it's less than that when black plays it smarter... ...in book.
1.e4 c6 2.d4,d6 i can say some wise words about, as i in fact tried this
with DIEP. It's complete walk over for white. Programs with a lot of space
play their best games usual.
these kind of openings focus upon sneaky pawn moves for black with which
you, after a mistake of white, can fight back by strategical means.
I can't remember anyone in world top humans use that dubious assumption
that opponent makes mistakes. Things like modern move order systems and
most pirc systems are of course objectively seen complete suicide.
Now there are 2 problems when you let a program play those type of
systems:
a) they lack the strategical knowledge to profit when it's possible,
and i doubt they ever will have that.
b) assuming both are strategical equally stupid, which is not a too
bad assumption, this means black is completely toasted as computer
doesn't make any mistakes within its own knowledge.
1.c3 of course directly goes out of book without advantage/disadvantage.
Well that's what you thought. How about letting a program try 1.c3 against
Diep at the world champs? My face will turn into a direct smile and the
smile will get bigger and bigger when my program is winning the game.
Best regards,
Vincent
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.