Re: [computer-go] KGS Tournaments: Format

[Heikki Levanto <heikki@xxxxxx>]

On Mon, Mar 07, 2005 at 01:52:22PM -0500, Don Dailey wrote:
> The reproducible program idea is not very good.  There is no way a
> multiprocessor program or a networked program could compete.  You
> should not require a program to be reproducible in order to compete.

If - and that is a big if - all programs gave out top-10 candidates for
moves, and possibly some reasoning behind them, then it should be
possible to estimate reproducibility even if the exact same moves have
not been chosen. This would still leave open a few ways to cheat (the
cheater could veto the selected move if the second-best would be better,
and both would show up in the top-10 lists)

I suspect there are programs that can not produce such a top-10 list
(although I have no idea why not), and/or that some participants might
hesitate to implement this feature, for lack of time, or for fear of
divulging too many of their secrets.

And we would still be open for endless discussions and accusations.

I suspect the best way is to trust people not to cheat, and reduce the
motivation by playing frequent tournaments with no prizes and not much
of bragging rights. In the long run, the cheater will loose, by pushing
the rest of us to improve our code, and by wasting his time with such
unproductive work as cheating.

-H


-- 
Heikki Levanto   "In Murphy We Turst"     heikki (at) lsd (dot) dk

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/