Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Unwanted consequences / side effects

Author: Robert Hyatt

Date: 10:27:22 04/17/01

Go up one level in this thread


On April 17, 2001 at 05:39:55, Ed Schröder wrote:

>On April 17, 2001 at 05:13:51, Rajen Gupta wrote:
>
>>On April 17, 2001 at 05:02:25, Ed Schröder wrote:
>>
>>>On April 16, 2001 at 19:53:23, Peter McKenzie wrote:
>>>
>>>>Its not that the programs can't play their best when the author is not present,
>>>>the point is that if the author is present we can be SURE that the program is
>>>>operating conditions which are optimal for it.
>>>>
>>>>If the program is operated by a 3rd party, then it is still possible that the
>>>>program is operating under optimal conditions BUT we have a number of issues:
>>>>
>>>>1) The 3rd party is unlikely to be as careful as the author.  The author has
>>>>invested many years in his program, and therefore has much greater motivation
>>>>for making sure that everything is set up 100% right.  All those little things
>>>>like selecting the right book, turning pondering on, configuring for the right
>>>>number of CPUs, setting the right hash sizes, making sure that no other
>>>>processes are stealing CPU, making sure tablebases are installed correctly etc
>>>>etc.
>>>>
>>>>Of course a 3rd party will probably get these things right, but if you had to
>>>>bet your life on it I think you'd rather have the program author doing it.
>>>>
>>>>2) The author will be much more capable of diagnosing any problems than a 3rd
>>>>party.  Problems?  What problems you say.  Well, lets think about hardware
>>>>problems for a start.  Memory can fail from time to time, and of course hard
>>>>drives can fail too.  And how about the CPU?  Remember when Ed Shroeder managed
>>>>to demonstrate that his Kryotech chip was faulty in one of the Rebel matches?
>>>>
>>>>These things are somewhat rare, although they are more common on the sort of
>>>>state of art hardware that is likely to be in use.  In any case, the author is
>>>>likely to spot the problem (and recommend a course of action) before anyone
>>>>else.
>>>>
>>>>3) Fairness: of course the 3rd party should be impartial, but how can we be sure
>>>>of this?  I don't personally know the people involved in the Kramnik
>>>>qualification match, although I assume they are probably fair and unbiased.
>>>>However I would have alot more faith in a competition being fair if the authors
>>>>were present because I know they are going to make sure that they are getting a
>>>>fair deal.
>>>>
>>>>Its like the old saying goes: 'Justice must not only be done, it must be seen to
>>>>be done'.
>>>>
>>>>That about sums it up really.
>>>>
>>>>Regards,
>>>>Peter McKenzie
>>>
>>>
>>>All good points Peter and of course a lot of people will agree with you.
>>>
>>>There is something else I would like to mention, an aspect that has not been
>>>discussed yet.
>>>
>>>There is an important chess event planned, very good. If the Braingames
>>>company would have said, "world champion vs world champion" than everything
>>>is okay with me, it is their show and their money.
>>>
>>>But the Braingame company said: no, we want the best program to play Kramnik
>>>and furthermore we hire a few experts to make that decision (Enrique/Bertil).
>>>
>>>So far so good.
>>>
>>>Then the 2 experts pick 4 programs based on THEIR OPINION.
>>>
>>>Right or wrong?
>>>
>>>Well, CCC is full of it. Lots of divided opinions.
>>>
>>>There are a few bad side effects I would like mention concerning this
>>>giant discussion:
>>>
>>>1) Whatever program in the end is chosen, it will not have a full public
>>>support.
>>>
>>>2) There is a small risk that the big division in opinions may lead that
>>>Braingames may decide to cancel the whole event as positive attention is
>>>certainly one of the financial aspects of Braingames to have this event
>>>and they certainly can not have bad publicity.
>>>
>>>This is in NOBODY's interest as I am pretty sure EVERYBODY here in CCC and
>>>elsewhere wants to have this match whatever candidate in the end is chosen.
>>>
>>>Therefore it would be wise this whole play-off thing should be reconsidered
>>>in such a way it has full public support, or at least a significant majority.
>>>
>>>Ed
>>
>>Hi ed: no method of testing is perfect but i dont see whats wrong with proposed
>>methd of selection.3 top of the range multi cpu progs(one a world champion,
>>another which has proved itself recently against a batch of A list grandmasters
>>and a third which has also proved itself repeatedly against top level humans and
>>heads the current SSDF list.)these will play a lrge series of matches gainst
>>each other to determine the challenger to kramnik. the people selected to do the
>>selection themselves are highly experienced as well as being known for their
>>personal integrity. i think we should encourage rather than try to ddestroy,
>>becuse as you rightly pointed out the sponsorrs might withdra any ttime!
>>
>>rajen
>
>
>"world champion vs world champion", it will have a big public support, don't you
>think?
>
>The planned "play-off" thing is just too debatable, CCC is entirely divided
>and for good reasons.
>
>My suggestions:
>
>1) Invite more programs, invite 3-4 GM's, play x human-comp games. Afterall
>the Braingames event is a human-comp event and not a comp-comp one, no?
>Authors are fully responsible.

This would be a reasonable idea.  However, I would suspect that putting it
together would cost a fortune.  ICCA events are not cheap and now you factor
in paying 4 GM players as well.  This probably won't happen as Kramnik is
going to want a big prize fund himself.



>
>2) Invite more programs and play a manual comp-comp tournament, authors are
>fully responsible.

Hmmm...  This has been done three times in the last 2 years.  Shredder won
_all three_.  The two WMCCC events and the WCCC event.  Again, this is going
to cost money.  Why do that when we _already_ have the results from the previous
three such events?




>
>I definitely prefer option 1 which IMO comes as closest to the truth playing
>Kramnik.
>
>I think both options will have a big public support.

Option 2 certainly has support.  Since it has already happened.  :)



>
>Ed



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.