[gnugo-devel] Fwd: [computer-go] April KGS on-line computer Go tournament

[David G Doshay]

Begin forwarded message:
> The fourteenth KGS computer Go tournament will be this Sunday, May  
> 7th, in the European evening, American morning, and Asian night.
>
> The Formal division will be on 13x13 boards, the Open division on  
> 9x9 boards.  Both will use fast time limits, 13 minutes each sudden  
> death and 8 minutes each sudden death respectively.




Hello GNU Go folks,

I would like to enter SlugGo in this tournament, and would like to  
know which of the divisions, if either or both, the GNU Go team  
prefers for itself. I will have enough CPUs to enter both, but the  
rules stipulate that I can only be in the Formal division with your  
permission. I think it best to apply the same logic to both divisions  
because of the way that SlugGo uses the GNU Go engine. That is to  
say, I will also only enter the Open division with your permission.  
Of course, a SlugGo win against GNU Go is not guaranteed, the cgos  
results show that SlugGo_l3 has about a 61% win rate over the ggexp  
versions, 71% over 3.7.4, and 83% against 3.6 (the version we are  
presently using) on 9x9.

We have not done any tests at all with SlugGo on 13x13, but are  
curious about how well we would do. There have been enough surprises  
on 9x9 that we have no basis for guessing how those results will  
scale to 13.

To be specific, SlugGo_l3 does 12 way branching at the top level, and  
then linear lookahead 3 more levels. SlugGo_88 does 8 way branching  
at each of the top 2 levels (64 CPUs doing lookahead) with the same  
maximum depth, and does slightly worse against almost all strong  
opponents, but most notable is the fact that the advantage l3 shows  
against GNU Go is completely erased. While this at first confused us,  
we came to realize that this is because with second-level branching  
SlugGo evaluated the mistakes that GNU Go would make as unlikely to  
happen because other choices were considered to be better, whereas  
with l3 GNU Go's mistakes were taken as the only likely response and  
thus were accepted and taken advantage of. It is not clear why 88  
played slightly weaker against other programs, where we had expected  
that the additional move choices might give us a better evaluation of  
a number of possible responses. But it is clear that the primary  
reason that the cgos ELO score of l3 is higher than that of 88 is  
that l3 takes proper advantage of GNU Go mistakes, while 88 out- 
thinks itself and rates those mistakes as unlikely. SlugGo_88 uses  
the same evaluation function for choosing the first level 8 choices  
as for the second level 8, and it is not clear to us why an  
evaluation function that shows some improvement when applied at the  
top level would be weaker against non-GNU Go opponents when applied  
to their possible responses. This is something we will investigate  
"soon."

But I digress ...

I would be happy to enter SlugGo in both divisions, but will accept  
the GNU Go team's decision with respect to each division.

Cheers,
David