[computer-go] KGS online Go Tournaments: two suggestions

[Don Dailey]

What I would like to see is something slow but not so extremely
slow.  Something like 11 hours so that a full game will be complete
in 22 hours - a new round begins every 24 hours.   

Programs like Viking, CrazyStone and Lazarus would play 9x9 very
strongly at these levels and 19x19 monte carlo programs would 
also play very well.

But games played over many days are kind of boring and I don't
know if people will want to tie up their computers for a single
game lasting so long.     A game that lasts 24 hours would be
fun to watch during the day.    I would suggest a system to
allow the programmer to make resign decisions.   Programmers 
can build this into their clients - a way to instruct their
programs to resign.   My programs of course resigns on its own
and so do some others.

- Don

 

   

 

On Mon, 2006-06-05 at 11:05 -0400, Weston Markham wrote:
> On 6/3/06, Aloril <aloril at iki.fi> wrote:
> > 1 move / day and 19x19 might be a bit too long tough unless something
> > like this is possible: I wonder if it would be possible to allow
> > updating bot to new version when *opponent* is thinking and has not yet
> > submitted move? This would allow to do 2 things:
> > 1) update bot when new version becomes available
> > 2) fix bugs that game uncovers (once those errors/blunders have been
> > made already on board ;-)
> >
> > However it would not allow to change code to make it play another move
> > when it is its *own* turn.
> >
> > I think this would make participating much more interesting for authors.
> > It also makes it much more challenging to be sure games are fair :-(
> > No free lunch ;-)
> 
> I do not think that this is such a good idea, at least as presented
> here.  Suppose I were to enter a program that plays its opening moves
> from a database.  Each day, I can simply add my own responses to all
> of the possible moves by its opponent.  The resulting sequence of
> programs plays exactly like me!  This is, of course, the extreme case.
>  However, the point is that competing in this manner might be too much
> biased toward tuning one's programs to the individual lines of play
> followed in the competition.  Consequently, it might not do much to
> advance the state-of-the-art of computer go in general.
> 
> Weston
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/