[gnugo-devel] Comments on tests

[Daniel Bump]

Trevor's views are indispensible for seeing what is going on
with the regressions.

I browsed Tristan's tests (except global.tst) and had some
comments. I didn't comment on ones Gunnar had already commented
on (via his patch) unless I had something to add.

capture test 4: The comment in Gunnar's patch (already in the CVS)
                shows that the recommended H5 may not work. But
                B:H6 might be the correct answer.

capture test 5: It looks to me as if S3 doesn't work. After
                W:S3 B:S2 W:T2 B:S1 W:O1 B:S5 black wins the semeai.

vie test 8    : Instinctively I want to play R7 which is GNU
                Go's move. However this leaves W the move at Q8
                so maybe R8 is best. I find it hard to believe
                the recommended S9 is good. B has a half eye
                at N7.

vie test 17   : GNU Go plays at E6 which probably lives though
                maybe E2 is better.

vie test 20   : The recommended answer H9 and GNU Go's Q14 are
                both unnecessary. B is alive as it stands. (J8
                and H9 are miai.)

vie test 27   : After GNU Go's answer at F3 it looks as if B is OK.

vie test 38   : GNU Go's answer at A4 works and is a little better.
                Interesting position.

vie test 39   : GNU Go's answer at D7 lives but is one point smaller
                yose. A replacement pattern would fix that (though
                not within the owl code.)

vie test 50   : Looks to me like this becomes ko after 
                B:G14 W:H15 B:G15 W:K17 B:J17 W:L16 B:J19 W:K19.

Dan