Re: [gnugo-devel] Life and death reading error

[Evan Berggren Daniel]

On Mon, 21 Apr 2003, SP LEE wrote:

> Dear all,
>
> Maybe this is a simple problem regarding life and death. In this game, the
> GNUGO's black group at upper right corner is actually dead, but GNUGO still
> thinks it's alive. I'm not quite good at reading the debugging information,
> but it seems GNUGO thought the group already has two eyes at P15 and T15,
> which seems to be rediculous to me.
>
> Does this mean the eye space data base should be tuned?

Actually, what is happening is that GNU Go believes there are two eyes in
the R13 / T15 space, because it does not see the cutting point.  If the
R12 string were solidly connected to the rest of the chain then black
would clearly be alive in seki.  If asked, GNU Go easily sees that T11
kills, however it does not try the move on its own.  I think this is
Paul's favorite problem, though I don't know what he's done with it.  It
certainly is a major issue, though.

Evan Daniel