Re: [gnugo-devel] Eye pattern 304

[Paul Pogonyshev]

SP LEE wrote:
> This is the game I posted before regarding life and death, where gnugo
> played black. After white played J2, the black group on lower side was
> dead. I think I'm not yet on the stage of fix any problem of the software
> (it's still too complicated for me), but after studied tons of debug output
> of gnugo I found some strange things. Before white played J2, I let gnugo
> analyze "owl_attack L2". It's good that it started to consider at first the
> attack position J2. For the following sequence it considered:
> W:J2 B:H2 W:J3 B:K2 W:J4 B:F2 W:E2 B:K1 W:L4 B:F1
>
> gnugo thought there was an eye at G1 of type 304 and the group was alive
> due to "getting deep, look lively". If I'm right, this eye is of type 0111.
> However, I think if attacker moves first, there should be no eye.

this is a real problem that needs some investigation (althought it's not
the main cause of owl's fail in this position).

pattern 304 is perfectly valid, but it shouldn't match in this position.
i'll try to look into the matter and determine why it got matched where
202 should have been.

Paul