[gnugo-devel] Re: gnugo-devel Digest, Vol 42, Issue 24

[John Tromp]

> > It next twice defeated 9x9 winner
> > CrazyStone but only after allowing gnugo to void a game in which
> > gnugo at lvl 15 took seemingly forever on a single move.
>
> Yeah, high levels tend to be unpredictable. Do you still have the
> position available?

It's the position at the end of gnu-crazy.sgf. I suspect GnuGo had
used more than
half an hour thinking time when I killed it.

> It's funny how this happens all the time. S11 was a really lousy move
> and T5 was at the very least unorthodox.

I think further improvements in GnuGo should first concentrate on understanding
these weird effects and making it robust against them.

> > But, it's the end result that counts, and we can congratulate all GnuGo
> > developers on this golden medal!
>
> Nice to be back on top, although this wasn't the strongest field the
> Olympiad has seen.

Ken Chen blamed the blunders made by his Go Intellect program on a rule
he had introduced to make plays for unsettled groups mandatory, but a bug
caused the rule to be misapplied. He also hadn't yet been able to track down
a bug in his gtp implementation, but hopes to have it working next time.

> > I will not be present at the prize giving ceremony tomorrow.
> > Please let me know to whom the medal should be sent.
>
> I wouldn't mind getting it. They should have my address from the
> subscription forms.

Ok, I'll tell the organizers.

regards,
-John