Re: [gnugo-devel] ld_owl tests

[Arend Bayer]

Evan wrote:

> I found this problem at
> http://senseis.xmp.net/?CompletelyArtificialProblem
>
> GNU Go has great difficulty with it, even in the later stages.  I thought
> it was interesting, and it doesn't look like there are any copyright
> issues, so here are some tests and an sgf.  The whole thing is probably
> too difficult for GNU Go right now, but the later stages should be
> solvable.

Well, if you _want_ to be pedantic, you can certainly find a copyright issue
here... (SL is under the open content license, which is incompatible with...)
But no, I don't think we need to worry about that here.

I am not sure how much good a completely artificial problem can do for our
owl test set. The pattern based shape approach for owl reading will always
be limited in the number of moves it can try, and we have to tune it to
be reasonably successful in the common cases.

The later stages of this problem should be solvable by the optics code,
however.

Arend