Re: Go program strength

[Nick Wedd <Nick@xxxxxxxxxxxxx>]

rbrown@xxxxxxxxxxxxxxxxx writes

Tim> There is also the theory that since the Computer Ing Cup and 
Tim> similar events are the main testing ground for go programs, the 
Tim> programs are designed to perform well there, rather than against 
Tim> humans. What do people here think about that? Is it just slander,
Tim> or is there some truth in it?
 
Nick> I think that there is truth in this.

Rich> Me too.
Rich>
Rich> Tim, I have to ask, why did you choose the word "slander"?  Do you 
Rich> think that programs should not be tuned to perform well against 
Rich> specific opponents?  Do you regard this as being somehow immoral?  
Rich> If it is true that programs in a competitive event are in fact 
Rich> tuned specifically to perform well in that event, do you think 
Rich> there is something shady about that approach?
Rich> 
Rich> I would regard this capability a desirable feature of a go-playing 
Rich> program.

There are two different points here.

1.  Do programmers tune their programs to be good at beating the
probable top contenders in these events?
2.  Do programmers tell their programs who they are currently playing,
so that they can adopt an appropriate style?

I believe that the answers are:  1. Yes.  2. No. 

But as applied to my own play, the answers are 
1.  No.  I try to play better Go.  I do not train myself specifically in
the tricks which might allow me to attain my next humble target, i.e.
promotion to shodan.
2.  Yes.  If I see that I am playing a certain Mr. Adams who has an
encyclopaedic knowledge of josekis, I will tend to open on 3-3 points,
and will certainly not risk a taisha.

So for question 1, programmers do things which I would regard as
shameful (though not unethical) if I applied them to my own training.  I
think that this may account for Tim's use of the word "slander".

Nick
-- 
Nick Wedd    nick@xxxxxxxxxxxxxxxxx