[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [computer-go] 2nd KGS Computer Go Tournament
X-Authenticated: #16844820
From: Paul Pogonyshev <pogonyshev@xxxxxxxxxxxxxxxxx>
Date: Mon, 9 May 2005 23:27:18 +0300
X-Y-GMX-Trusted: 0
X-Spam-Score: -4.9
X-Spam-Flag: NO
X-Scanned-By: MIMEDefang 2.42
X-MIME-Autoconverted: from quoted-printable to 8bit by theory.csail.mit.edu id j49KJ7OS015983
Don Dailey wrote:
> > I do not think anyone will try to add dispute-negotiation code to their
> > bot. But if anyone is daft enough to do this, I shall insist that they
> > keep it switched off for any event that I am organising.
>
> But I have this in botnoid and it was turned ON during the tournament!
Just wondering: how does it look like? How can you do any dispute
negotiation over GTP if it is designed so that engines have no direct
control over what is happening whatsoever?
Paul
I'm glad you asked.
The KGS protocol DOES in fact allow some limited control.
But first of all, let me mention that "final_status dead" and
"kgs-genmove_cleanup" IS a real disagreement resolution protocol.
Your program has a measure of control for a few reasons:
1. You can report a false number of dead stones.
2. You have a choice of either ignoring the kgs-genmove_cleanup
protocol or accepting it.
Point 2 may not seem important, but having a choice is at least as
good as having NO choice and gives you some control.
For instance, ignoring kgs-genmove_cleanup give the program the
ability to pass a decision to the tournament organizers. I thinks
it's possible to make this work to the advantage of the program. I'm
thinking in terms of seki detection or at least human assisted help in
this regard. So it may be possible for the program to manipulate the
tournament organizers into resolving seki, when a mistake could be
fatal to the program.
But the fact that the protocol is optional should be able to be used
even more to your advantage. I think for my program, it's better to
let the tournament organizer make the tough decisions but I'm still
thinking about this.
- Don
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/