Re: [gnugo-devel] current regression failures

[Arend Bayer]

> In the detailed listing below, the module within parenthesis is the
> one which was involved in the change in status. I'd appreciate if
> stronger players could review my questions and comments about the test
> cases themselves.
Although I am maybe not much stronger than you (I am 3D), here are
my comments:

> atari_atari:1 (atari_atari)
> Correct answers at L3 and N2 are missing.
Agreed.

> neurogo:16 (atari_atari?)
> I don't understand this test case. 
Neither do I.

> strategy:27 (optics?)
> Compared to 3.1.15, an owl attack on D2 is found. This together with
> the fact that the proposed owl defense of D2 is bad (F2 rather than
> C3) causes the failure.
The owl attack seems to be wrong. Still, C3 should be found rather than F2
by owl as well; owl does not even try C3. This seems to indicate a
missing pattern? A workaround that might solve other problems as well
would be to introduce a pattern that gives C3 a c classification.

> strategy2:73 (semeai, moyo?)
> To me it seems like E7 is very similar to F7. Any reason not to add it
> to the correct answers?
I think E7 does not capture the black G11 group: W E7 - B F8 - W H7 - 
B F7 and F6 and J7 are miai. It took me quite some time to spot this
when I first came across this problem.
This got indeed broken by my moyo patch. H16 dragon had moyo size 0 before
my patch and has moyo size 12 with my patch; this means the strategic
attack value of R17 is less than before. This test should break anyway once
the duplication of strategical attack/owl attack threat is eliminated; GNU
Go does not think that R17 kills so it should not play it. Btw, this
might be a good (though challenging) example of an owl attack with cut
threat.

> nicklas5:804 (moyo)
> This is a case of several bad strategic values. That it passed before
> and fails now is mostly coincidental.
Another example of the cautious strategic evaluation of owl-attacks causing
problems. Also, H6 gets too much strategical attack value for the dragon
E6.

> trevor:381 (atari_atari)
> This actually passed before due to a bug in the atari_atari code. Now
> the main problem is with an incorrect reverse followup value.
Indeed there seems to be a duplication. OTOH, I think E8 is fine and
should be added to the correct answers.

> global:26 (moyo)
> Very shaky valuation. Mostly luck it passed before.
H3 and H4 are typical examples of moves whose overvaluation should
be reduced with a continuos territorial evaluation. OTOH, GNU Go currently
does not see how much can be achieved with D12.

> trevora:120 (atari_atari)
> Various valuation mistakes. However, the previous move at E17 isn't
> satisfactory and shouldn't be listed as correct. Or am I missing
> something?
I come to yet another conclusion :-)
I think that B8 is the correct move. W can answer E7 with E8 and then D8
with D9. Compared to D8, this only gives up two points around D8, yet
B8 secures B7 and A6 (not B5 due to a shortage of liberties, but B5 cannot
be saved anyway unless I am missing s.th. again).

> strategy4:166 (atari_atari)
> Mistake by atari_atari. Returning multiple defense moves best
> solution.
Wow, incident94 is really a funny game!

> nicklas1:1106
> What's so bad about F6? 1 point reverse sente is maybe not the biggest
> move on the board but it's also far from the worst. This test should
> either be removed or changed to require the largest move on the board
> (I guess this would be Q6 or some move around T12).
Largest is probably the sequence Q6 - R7 - R5 - Q4 - Q4 - S5
(not Q6 - R7 - R5 - S5? - Q4!). This whole sequence is probably a bit too
subtle for GNU Go at the moment. Still, we should try to find a way to
value 1pt sente moves like Q6. Also, 3.3 strategical value of F6 for
the P8 dragon is somewhat weird.

> owl:136
> I don't understand this test. Supposedly F14 would kill but black
> responds at B12 and has two certain eyes.
Seems Trevor has already taken care of this one.

-Arend