Re: [gnugo-devel] Semeai weirdness.

[bump]

>Part of the complexity of the function is caused by the fact that a dragon
> may be involved in several semeais. We need to collect results for the
> different semeai, analyze this information, then update statuses.

Another point is that this information is not really sufficient
to update the statuses correctly.

Suppose that dragon A is involved in semeais against dragon B and
dragon C and that both semeai are critical.

Is dragon A alive, dead or critical?

All three cases are possible. If A wins the semeai with B
and A is to move, yet if it takes a lot of moves to remove B
from the board, A may lose the second semeai and A is dead.
On the other hand if B has fewer liberties, A may be alive.

The function newsemeai attempts to draw a conclusion here,
yet no correct conclusion is possible.

One thing that is urgently needed (for other purposes, even
when there is only one semeai on the board) but not provided
by owl_analyze_semeai is an estimate of how close a semeai
is. Yet even that information may not suffice to resolve
this question.

Dan