[Computer-go] I need an off-the-shelf final position live/dead evaluator

Gunnar Farneb├Ąck gunnar at lysator.liu.se
Mon Nov 29 15:24:59 PST 2010


On 11/28/10 10:15, Dave Dyer wrote:
> At 10:39 PM 11/27/2010, David Fotland wrote:
>> Accurate scoring, even at the end of a game, is very difficult.  You have to
>> read accurately, and evaluate semeai and seki.
>
> Yup.  I spent years developing the capability to score endgames
> at the point where humans typically leave them.  Getting within
> a few points of correct 95% of the time is achievable.  The other
> 5% you will either make whopper mistakes or never terminate.
>
> This is now pretty old, but I don't know of any more recent or better
> results. http://www.andromeda.com/people/ddyer/go/scoring-games.html

On that testset GNU Go 3.8 finds the correct score in 85.6% of the
cases, is off by at most one point in 98.2% and within 3 points in
99.4%.

And well, it can be argued whether humans in general (as opposed to go
pros) consider the game finished and ready to be scored with open kos
remaining.

/Gunnar



More information about the Computer-go mailing list