Re: [computer-go] future KGS Computer Go Tournaments - two sections?

["Eric Boesch" <ericboesch@xxxxxxxxxxx>]

I think that David Doshay's evil-twin effect (an advantage over an opponent 
that is in part not a consequence of general playing strength, but of 
knowing what your opponent is going to do, in particular because you share 
that opponent's code) does not belong in serious, top-tier tournaments.  
Since David forthrightly claims that his program exhibits that effect with 
respect to GnuGo, I think the two programs do not belong in the same serious 
tournament, despite the considerable enhancements that David's group has 
made.  (Daniel Bump's suggestion to potentially have GnuGo sit out a 
tournament in favor of SlugGo is an excellent solution to this problem.)

Suppose KillerGo scores 1-1 against GnuGo-based GnuGoKiller, 4-2 against 
other GnuGo variants, and 8-0 against the rest of the field, whereas 
GnuGoKiller goes 1-1 against KillerGo, 6-0 against other GnuGo variants 
(possibly because of the evil-twin effect), and 6-2 against the rest of the 
field.  If I believe in the evil-twin effect, I will say that KillerGo is 
probably the best program and that GnuGoKiller may not deserve the share of 
first place that it got by beating up on its weaker cousins.

Less serious tournaments are a different matter -- there, I think SlugGo, 
GnuGo, and GoFigure should all be allowed to enter simultaneously.  There's 
a lower effort limit -- tweaking a few parameters doesn't make for a new 
program -- but SlugGo is certainly a lot more than that, and I assume 
GoFigure is also.  Most of us would like to see them play, and anyone who is 
too bothered by the evil-twin effect, which I suspect is not large anyhow, 
can enter the more serious tournament instead, assuming their program is 
good enough to deserve to be taken that seriously.

Daftgo is a special case -- it's a deliberately weak, non-serious GnuGo 
clone intended as a sparring partner for beginners.  Personally, I think it 
might be interesting to throw a fixed 30-kyu level program like Daftgo and 
maybe a 20-kyu level program as well into a 'B' level computer tournament as 
benchmarks, but I have no strong feeling about it either way.

I concur with the general feeling that multiple programs by the same author, 
group of closely collaborating authors, or commercial entity do not belong 
in a formal tournament, and generally not in informal tournaments either.  
Requiring a short description (usually a URL) in some language or other as 
part of computer-go tournament entries would help the referee to determine 
whether entries are different enough, while also adding interesting reading 
to the tournament crosstable.

Eric

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

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