[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: unmake move?



No.

You have to distinguish between full board search and
string tactics.

Many Faces only does about 20 nps in full board search.
That's because each full board evaluation includes many
tactical searches to evaluate tactical stability of strings 
and connections.

Most of the strong programs have slow evaluations because you
can't evaluate the score without figuring out the strength
of groups.  And you have to do tactical searches to figure
out if groups are stable.

String tactics is fast since the evaluation is just the
number of liberties.  Stable is 5 liberties, and captured is
zero.

I can't imagine a strong program that doesn't do tactical
search during full board evaluation.  I've seen many
programs that tried to avoid tactical search, but they were
all very weak.

David


At 03:23 PM 6/8/00 -0400, Don Dailey wrote:
>
>
>> 300K nps is very good.  You must have a very simple evaluation
>> function, and a very fast move sorter. 
>> 
>> Most of the strong go programs get less than 100 nps, since the
>> evaluation is so slow and complex.  With very slow evaluations,
>> unmake time is not an issue.
>> 
>> Many Face's tactical analyzer gets about 50,000 nps on a fast PC.
>> Its time is dominated by move selection and sorting.  Evaluation
>> for string tactics is trivial (just count liberties).
>
>
>
>P.S.  And that's what amazes me.  Your program is one of the best,
>      and it does about 50,000 nps, and yet other very good
>      programs do only 100 nps.  My program does 300,000 nodes
>      per second but stinks!   It shows the complexity of the 
>      game and demonstrates that there is much room for imagination
>      and varied approaches to writing a quality program. 
>
>