Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Shredder wins in Graz after controversy

Author: Sune Fischer

Date: 14:52:42 12/09/03

Go up one level in this thread


On December 09, 2003 at 14:00:08, martin fierz wrote:

>>Are you suggesting the UCI protocol should be forbidden? :)
>
>not at all. if your engine claims a draw in a position where it might still win,
>that is your problem. i'm saying that you can gain something by not
>automatically claiming a draw whenever you can. as i said, if every UCI GUI
>really claims a draw on 3fold repetition, then why didn't the shredder GUI claim
>a draw by repetition after it occurred?? as you can see, it gained something by
>not doing it :-)

Well if the engine thinks it can still win it shouldn't want to go into the
draw.

Not claiming a draw by 3 fold repetition (by either GUI or engine) I'd consider
a bug.

The problem is the game can theoreticly go on forever.
Imagine starting a game and night and watching it in the morning at move 2314
where the last 2250 moves have been repetitions.
That is the kind of senario you will face if you don't enforce it as a rule in
computer games.

>
>>Not really.
>>The engine always plays the best moves it can find, if that is a repetition then
>>it's a repetition. No point in continuing because the engines sees the draw and
>>wants it.
>
>not at all. the evaluation of a position where you have a draw in hand should
>not be 0.00 but ">=0.00". that's how we humans think about such positions. the
>programs are just too stupid to understand that :-)

Well yes, but if you want a draw it's because the score is <=0.00 :)
Of course the score returned by the search is an upperbound and if it is 0.00
then you have nothing better than a draw.
At least the program thinks it has nothing better than a draw, but it is also
the program playing... :)

>you can and should have your opponent show you that he knows that it's a draw,
>if you have a safe draw. for humans this concept is obvious. i can't help it
>that it's not obvious to an alpha-beta tree searching program!

It doesn't happen often that they play out a draw by accident, I guess because
the draw window is zero, any fluctuation will make the winning side play
something else.


-S.
>cheers
>  amrtin



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.