Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fritz5 cooking at SSDF and Nunn test set

Author: Robert Hyatt

Date: 08:20:34 02/26/99

Go up one level in this thread


On February 25, 1999 at 08:35:07, Vincent Diepeveen wrote:

>On February 25, 1999 at 08:06:55, Robert Hyatt wrote:
>
>>On February 25, 1999 at 07:09:26, Vincent Diepeveen wrote:
>>
>>>On February 24, 1999 at 22:06:34, Robert Hyatt wrote:
>>>
>>>>On February 24, 1999 at 17:48:10, Vincent Diepeveen wrote:
>>>>
>>>>>On February 24, 1999 at 16:04:41, Dann Corbit wrote:
>>>>>
>>>>>>I have an email from Frederic Friedel, and I am fully convinced that Fritz did
>>>>>>not 'cook' anything to get the answers right.
>>>>>
>>>>>I got the same email from Frederic and i'm completely convinced that Fritz
>>>>>did cook.
>>>>>
>>>>>Frederic is playing the innocence himselve, but in the meantime he has
>>>>>ordered to make an auto232 player for fritz that doesn't allow rebel9
>>>>>to learn, that exchanges colors so that other learners of other chessprograms
>>>>>get confused, and that's just the top of the iceberg. This top has
>>>>>been confirmed by Karlsson, which i honour for being so nice to admit
>>>>>that fritz autoplayer doesn't allow learning.
>>>>>
>>>>
>>>>I disagree with the above.  The 'shortcoming' is in Rebel, _not_ in Fritz.
>>>>IE if you can't learn when alternating colors, what good is learning, since
>>>>_most_ real tournaments do alternate?
>>>
>>>>I've said it before.. what you can fix on _your_ end you should.  In this
>>>>case, it is a strange design decision indeed that says you have to play N
>>>>games same color to learn anything...
>>>
>>>Comon get real Bob,
>>>
>>>You make a protocol, and i'll invent something to cheat with it.
>>>
>>>Example when playing crafty using winboard protocol one can easy
>>>set up a position every move, and prevent crafty from learning then.
>>>
>>
>>Nope.. 'position learning' still works so it still won't 'repeat' losing
>>lines forever.
>
>Noop. you can't write your learning files. See below
>
>>>For every protocol one can invent such things. We have a protocol to
>>>play each other, now unless the protocol is the protocol of a fool,
>>>we can expect that we use a chessgame to fight, and not the protocol.
>>>
>>>For every protocol you make i can make my own autoplayer that prevents
>>>you from learning!
>
>
>>I'll take that bet...  this is just 'incomplete programming'.  The auto232
>>protocol doesn't allow one side to prevent the other from learning.  If the
>>program requires some 'key' from auto232 to 'learn' that is a bad design.  If
>>the program depends on N games with the same color, that is a bad design.  But
>>I'll bet you can play crafty all the auto232 games you want it _it_ won't fail
>>to learn whatever you do.  Even if you hit ^C to terminate it in the middle of
>>a game, it will 'learn'.
>
>Before we start playing i simply writeprotect your directory.
>
>As in my autoplayer the protocol says that you must handle that yourselve.
>Of course crafty gets killed after a game
>
>Then it gets reloaded. I will repeat my game...

I don't buy that as a 'fix'.  Because if you try it in crafty, you'll find
it simply plays _without_ the book.  Even worse, you will find it crashes
before doing _anything_.  And that would not lead to a game.

And there is no way to protect against 'external' stuff like that, because you
could also use a debugger and modify my executable to play differently by
zeroing a few things here and there.

We were talking about the _engine-to-engine_ protocol.  _not_ about what the
operator could do externally.

But in any case, you write protect my directory, I cant play.  The book files
won't open (I have w+ because of learning), the position learning files won't
open (which will crash me) and the log files won't open (same result).




>
>And if this doesn't work then i try things at home till i find a gap
>in crafty. If i have found this gap then i will consirably score better
>against it.
>


no protection from that, except you will only get away with it for one
game, before it 'learns' not to play that opening line again.



>>>This is what happens here.
>
>>>Ed says suddenly something else now, meaning that more games it didn't
>>>learn, but from Karlsson i received confirmation that Fritz doesn't
>>>send the 'save the game' command.
>>
>>
>>So I should write code that depends on "my opponent" to trigger my learning?
>>
>>As they say in the Navy, "Not on _my_ watch."
>
>Well the watch always forgets that he has written down that he must watch,
>so he'll not watch.
>
>>>
>>>Now i don't know how fritz5.16 represents results on the screen after an
>>>auto232 session as this autoplayer is secret, so we can expect something
>>>that suits chessbase.
>>>
>>>>>Even more important than disallowing learning is the fact that only fritz5
>>>>>auto232 player can collect the results.
>>>>>
>>>>>Let me give an example what happened.
>>>>>
>>>>>I got about 6 months ago a call from a tester who played around 20 games at
>>>>>auto232 player against Rebel9.
>>>>>
>>>>>He told me that i should stop chessprogram development, as my program
>>>>>didn't win a single game versus rebel9.
>>>>>
>>>>>So i first asked whether all games were more or less the same (learning
>>>>>from rebel9 you never know...). He told me clearly that the games were
>>>>>not the same as he turned off learning.
>>>>>
>>>>>This confused me. How can one not win a SINGLE Game vs rebel9?
>>>>>It was quickly solved when i got back the games.
>>>>>
>>>>>Diep won lucky several games, but those games were long. The games it
>>>>>lost were short (short after book). Diep was mated within say 50 moves
>>>>>or something. The games diep won were all 60 moves, after which the
>>>>>auto232 game stopped the games. However none of the games diep had
>>>>>mated rebel9. Further a big bunch of the games were clear draws, or
>>>>>3 fold repetition.
>>>>>
>>>>>Now diep doesn't collect game results. Rebel does, and the WAY in
>>>>>which it does caused the confusion that diep didn't win a single game:
>>>>>
>>>>>rebel9 screen showed a questionmark where diep won the game,
>>>>>and some draws got a questionmark too. Where rebel9 was winning, or
>>>>>had mated diep, it showed that rebel won.
>>>>>
>>>>>People that auto232 play are usually very very lazy, and they have
>>>>>the right to be so. So they simply see what is at the screen and
>>>>>pass that on.
>>>>>
>>>>>Now this is i think very important to realize. ONLY FRITZ5 CAN SHOW
>>>>>THE RESULT OF THE GAME, which might give the same confusion.
>>>>>
>>>>>Every new game that gets started the opponents learning is fooled
>>>>>(if it has learning), and the number of games played is very inconsequent.
>>>>>
>>>>>Please check out the games of fritz5 versus mchesspro and see what happened.
>>>>>So don't look to the ENDRESULT of it, but do it game by game.
>>>>>
>>>>>For example:
>>>>>After the first game it was 1-0, after the second game it was 1-1.
>>>>>Please do that for all games, so if 40 games are played i expect
>>>>>a graph of 40 dots, also write down what openings line was played,
>>>>>and some whizzkids here might slowly see how
>>>>>silly SSDF playing is, and what the influence of a topdown form
>>>>>of learning is when you play too much games with 2 small books.
>>>>>
>>>>>Yes DIEP is not yet at SSDF, because its learning doesn't work yet.
>>>>>Without learning you can go home.
>>>>>
>>>>>Now this naive email from Frederic to me and to corbit is kind of weird.
>>>>>
>>>>>He as the boss must have known everything. His assumption that fritz
>>>>>plays simply very good in the opening when playing on its own, is
>>>>>laughable. That doesn't explain the difference of playing at the
>>>>>nunn test set from 5.16 versus 5.32
>>>>>
>>>>>Please post an open email about that here Frederic. I know you read this.
>>>>>
>>>>>If you are already reading that, then tell us why you have ordered
>>>>>to make your own auto232 player special for SSDF, instead of a 20
>>>>>minutes implementation of the donninger autoplayer. Was beating
>>>>>programs more important than 'how' to beat them? Was beating them, taking
>>>>>advantage of the protocol and the fact that rebel learns only after
>>>>>it stores a game, the only way to beat them?
>>>>>
>>>>>We hear that you too do a lot of efforts to play as well as fritz5
>>>>>can do. If i were in your shoes i would get myselve a good bookmaker,
>>>>>the best one money would be able to buy.
>>>>>
>>>>>You better prepare, i plan to kick butt of fritz5 in world champs!
>>>>>
>>>>>Just like Bob i've prepared in a different way. I plan to run parallel!
>>>>>But don't worry, even a quad xeon system gives diep 60k nodes a second,
>>>>>where fritz gets at a PII-450 already a quarter of a million nodes a second
>>>>>or something.
>>>>>
>>>>>Vincent Diepeveen



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.