Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Updating engines during tournaments? (Odyssee Tournament)

Author: Chessfun

Date: 13:51:28 03/06/01

Go up one level in this thread


On March 06, 2001 at 16:13:12, Robert Hyatt wrote:

>On March 06, 2001 at 11:46:36, Brian Kostick wrote:
>
>>On March 06, 2001 at 10:53:37, Robert Hyatt wrote:
>>
>>>On March 05, 2001 at 22:48:59, Brian Kostick wrote:
>>>
>>>>On March 05, 2001 at 16:08:48, Andreas Schwartmann wrote:
>>>>
>>>>>On March 04, 2001 at 11:51:18, Thorsten Czub wrote:
>>>>>
>>>>>>On March 04, 2001 at 10:36:15, Ulrich Tuerke wrote:
>>>>>>
>>>>>>>Die B.27 ist okay, unterscheidet sich kaum von der Paderborn-Version.
>>>>>>>Gruss, Uli
>>>>>>
>>>>>>
>>>>>>brilliant. i ask because i don't want that anybody feels disadvantaged.
>>>>>>if anybody thinks he has a better version he is allowed to upgrade
>>>>>>between the rounds.
>>>>>
>>>>>
>>>>>And that's completely rubbish. If you update engines between rounds, what use is
>>>>>the outcome of such a tournament? There is no consistency in this tourney! An
>>>>>updated engine is a DIFFERENT engine, so you might as well not call it a tourney
>>>>>but a set of engine matches. Hell, you might even start such a "tourney" with
>>>>>Fritz 1 and end up with Fritz 7 ... and what would this say about Fritz's
>>>>>playing strength? He started weak but ended up the winner nontheless? Har har.
>>>>>In my opinon, the engine version that started the tournament should be the very
>>>>>engine that ends it. No changing of horses in midstream or else the results get
>>>>>worthless!
>>>>>
>>>>>Imagine Linares ... Kasparov gets bored in midtourney and gets exchanges by
>>>>>Kramnik ... Shirov does not play to good, so he sends in his brother (does he
>>>>>have one?) ... but that would not be a tourney anymore. Just like your Odyssee
>>>>>with updated engines is no tourney in my opinion.
>>>>>
>>>>>Just my $0.02.
>>>>>
>>>>>Any comments?
>>>>>
>>>>>Andreas
>>>>>
>>>>>
>>>>>
>>>>>www.andreas-schwartmann.de
>>>>
>>>>
>>>>  I think I understand your point. I also note that some are objecting, finding
>>>>no fault with changes between rounds.
>>>>
>>>>  For an example let's say I am to tout Nimzo. Now I may want several tuned
>>>>books, anti-Fritz, anti-Crafty, anti-Whatever. Also I can tuned Nimzo engines
>>>>parameters, if fact use several versions of Nimzo. (99, 2000, 7.32, 8, ect...)
>>>>Its author may even go so far as to change code and recompile between rounds?
>>>>After all this it's fair play? No one objects? Obviously some DO object. Not to
>>>>how an individual host a tournament, but to how the results maybe presented.
>>>>
>>>>  All these changes designed to selectively defeat opponents and claim the glory
>>>>(customer base, money, rating, pride, ect...)? I personally would find it a
>>>>warped glory if it was not the identical product that I released for public
>>>>consumption.
>>>
>>>
>>>Then warped glory is what you get.  Because it happens all the time.  When we
>>>won our second World Computer Chess Championship with Cray Blitz, we made a
>>>significant change after round 2 where we lost a game we should not have lost
>>>due to a bad change I made prior to the tournament's start.
>>>
>>>_every_ computer chess event I have attended has had programmers making changes
>>>between rounds.  From something as simple as a different book to avoid repeating
>>>an opening you just lost in the last round, to adding some analysis for a book
>>>line your next-round opponent seems to want to play given the chance.  The
>>>changes might be more complex modifications to source code to fix a bug or
>>>cover up a hole you found in the last round.  You might find something in your
>>>time allocation code you didn't like and change it to behave more like you think
>>>it should.
>>>
>>>Does that mean all the WCCC/WMCCC/ACM event results are bogus?  I don't
>>>think so.  I have played in many _human_ tournaments myself.  And I often spent
>>>a lot of time modifying my _own_ book to get an advantage against a known
>>>opponent.  Humans do this all the time.  Is _that_ also bogus?
>>>
>>>
>>>
>>>>
>>>> IMO, such between round changes make for warped and useless results except to
>>>>maybe some engine authors or a few select others. Definitely not what many of us
>>>>expect when we look at tournament results and try to form some impression or
>>>>conclusion. Regards, Brian K.
>>>
>>>I would say at least it is no worse than playing on with an older version that
>>>has had a known bug fixed in a newer version.  Why should a program have to
>>>suffer on with a known problem after it has been fixed?  Commercial programs
>>>release bug fixes as well.  Should _those_ also not be used?
>>
>>  It is not for you to say that I have to accept someone elses warped glory. I
>>maintain the right to be selective. We've seen one major computer chess game
>>accept lowered status due to defective time control. Their representative said
>>something similar to: We accept this, it's our fault the time control does not
>>work properly. A gracious acceptance of reality. Yet when the patch came out of
>>course we look foreward to future tournaments with better standing.
>
>I didn't say you had to accept anything.  By your definition, you will be
>forced to ignore the results in each of the WCCC events, each of the WMCCC
>events, each of the older ACM events, each of the paderborn events, each of
>the many other events where computers played but had their authors present and
>able to make adjustments to the book or the program between rounds.  Note that
>that is _every_ major computer event ever held.  And you just invalidated every
>one of them.
>
>But continuing, DB was altered a bit after round 2 of the Kasparov match 2.
>Rebel was not the same rebel used in every one of the GM challenge games that
>Ed played.  The SSDF has allowed bugfixes or versions slightly newer than the
>versions available commercially.  They even allowed non-commercially-available
>versions of commercial programs (the famous hidden autoplayer discussion).
>
>If you use your reasoning, you have _no_ data whatsoever to go on except for
>the 'basement tournaments' where someone buys a copy of several programs and
>plays them in a tournament "unchanged" between rounds.
>
>
>
>
>>
>>  I will view tounament results that are more representative of the user
>>product. i.e. Chessfun's results, SSDF, ect... I maintain: If you go to
>>tournament unprepared, you lose. As you are fond of pointing out: We agree to
>>disagree. With respect for your chess programming, Brian K.
>
>
>You have already disqualified SSDF because they have done exactly what you don't
>want to allow, on a few occasions.

I believe on those few occasions the effect to be minimal as they still
use the release and only update as a patch is available for all consumers.

In the cases I remember the only one that had any amount of games was Junior 6
prior to the release of the A version. The proportion of games is likely no more
than 10% and the effect on it's rating no more than 10 points as I recall Uri
and I wrote on this subject last year.

Sarah.








I would also hope that you don't think the
>"same" Kasparov is playing in every tournament he enters.  Kasparov is evolving
>just like the chess computers are.  To compare his play of today with his play
>of 20 years ago would show just how far he has come.  To pick _any_ single event
>and say "this represents _the_ Kasparov" would be wrong.
>
>Crafty is Crafty... whether it is version 18.3 or 18.4.  Just because version
>18.4 fixed a small bug in 18.3, does not mean that if you start with 18.3 that
>you should stick with it.  Humans self-correct bugs in their chess playing all
>the time.  At least the ones that are getting better do.  I don't see why this
>should be an issue at all since every available computer event over the last
>30 years has been run in _exactly_ the way you don't like.  Personally I have
>been guilty of only a few between-the-round changes.  Chances are you add more
>bugs than you remove with so little time to write and test the changes.  But I
>have sat around in the tournament hall and talked to programmer after programmer
>as he madly made changes to fix problems he saw in the previous round...



This page took 0.01 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.