[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [computer-go] Third KGS tournament: game-end protocol
On 9, Jun 2005, at 2:20 PM, Don Dailey wrote:
I really do not believe it is in the interest of computer
Go to require that all dead stones be removed. It leads to end game
play like that of botnoid, and a check of the comments near the end
of
the SlugGo botnoid game show that while it can be temporarily funny,
even that gets old fast.
Why were you annoyed?
Because it brought out an unnecessary bout of snideness from the
observers.
Your program was the one that was winning and I
am the one who had to sit through the lost game and the rude comments.
How unpleasant could it have been for you?
It had nothing to do with winning and loosing. From my perspective
we are debating a protocol that will result in a specific end-game
behavior. I think botnoid played its end-game in a manner very true
to the result we will be getting from our bots if the simple protocol
being discussed is chosen, and I see how spectators react.
I don't want that snide attitude to spread over the results of all our
work. This is too hard for that. You work to hard on it and so do I.
You should not have had to sit and listen to that noise from folks
who have not tried to program this game. But you did, and all I am
saying is that the decided upon protocol should not encourage
our programs to play an end-game that is subject to derision.
From my very first game of Go I was impressed by the end game where
both players players pass when there is nothing more to be gained.
When a dispute happened, usually because one player was more
inexperienced than the other, the situation in question would be played
out. It was always specified where the dispute on the board was, the
stones in question are pointed to. All I want is for the protocol do the
same, and along with the "please consider playing this out" request
is a list of the stones in question.
Let's be reasonable. Anyone who writes a computer program to compete
is probably going to focus on playing strength first, and then the
niceties like ending the game cleanly and early. A protocol to do
this is a positive thing. But like the game of GO itself, you can't
force players to pass if they don't want to or are not very good at
it.
Indeed, no one should be required to pass, and there should be no rule
that requires. or encourages under risk of loosing, a program to play
significantly different than a person should. A program should not be
able to demand that its opponent play first or loose.
Yes, younger programs will have less polish than older ones. Our
programs are young and the subtile details will evolve slowly and at
different times in each program.
But please lets not accept a protocol that strongly encourages our
programs to play to the last liberty over the whole board. People
will stop playing against bots if we make that mistake.
David
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/