Re: [gnugo-devel] Regression results for the large scale patch

[SP LEE]

>
>
> Here are the regression results (against 3.4) for the large
> scale capture patch posted in
> <http://mail.gnu.org/archive/html/gnugo-devel/2003-07/msg00209.html>
>
>
>
> In total, there are 13 PASS (two more if we include the
> positions shown in the original message) and 19 FAIL. The
> fails and some of the passes are discussed below.
>
> I think that 6 of the fails are real, 8 can be interpreted
> as reliable alternate answers, while 5 are too difficult
> for me.
>
> Comments are most welcome, of course.
>
> Stephane.
>
>
> **********
>

I have put some comments below some problems.

>
> 39 unexpected FAIL: Correct 'R7', got 'R6'
>     http://www.public32.com/regress/?nngs1:39
>
>     Maybe not a fail. The engine believes that
>     R6 kills R3 while protecting the territory

Black R6 makes white S4 to sente and the white group in the corner will
live in ko.
Black R6 seems to give white no chance to live.

>
> 21 unexpected FAIL: Correct 'S16', got 'S14'
>     http://www.public32.com/regress/?neurogo:21
>
>     Real fail. S16 has more effect on the corner
>
>
> 2 unexpected FAIL: Correct 'Q18|S16', got 'P16'
>     http://www.public32.com/regress/?arion:2
>
>     Don't know. If P16 indeed kills R17,
>     then it's OK. I'm not strong enough to tell.
>

P16 indeed gives white too much chance to live. For example, white S15.

>
> 13 unexpected FAIL: Correct 'R13|M8|L9', got 'J3'
>     http://www.public32.com/regress/?lazarus:13
>
>     Maybe not a fail. J3 threatens to kill the
>     group at the bottom.
>

J3 is indeed a threat, but after white answers at J2, black has to play
at H3, otherwise there is too much aji. Then white has chance to escape
the group in the middle. So J3 is not absolute sente. Besides, we have
to check if later black will kill the middle group for sure or not.
Actually I like to play L8.

>
> 64 unexpected FAIL: Correct 'B13|C13|C12|E14|D14', got 'B12'
>     http://www.public32.com/regress/?strategy2:64
>
>     Maybe not a fail. Doesn't seem different from B13.
>
I think B12 is better than B13 in this case. For B13, black C12 is
rather naughty.

>
> 89 unexpected FAIL: Correct '(D16|B7|G13)', got 'H11'
>     http://www.public32.com/regress/?strategy2:89
>
>     Maybe not a fail. H11 has more effect in the center, but
>     leaves more freedom for the white stones.
>

If white got more freedom, then the black group E16 will be endangered.
I think H11 is not urgent.

>
> 440 unexpected FAIL: Correct 'O2', got 'N3'
>     http://www.public32.com/regress/?nngs:440
>
>     Don't know. O2 captures the corner more tightly,
>     but N3 sort of protects the cut in O4.
>

N3 won't capture the corner.

I didn't check all the problems. I think it's better to call owl_defend
again for the largely captured group to see it it is captured for sure.
In most cases, the result of owl_defend is dependable.

SP Lee