[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnugo-devel] Owl patterns
[gnugo-devel] Owl patterns
Sat, 11 May 2002 21:29:59 -0700
> Also, there are a number of owl problems that are tough to tune, because
> it's not obvious how to cut the nodes to under 1000. Increasing the node
> limit to 2000 may make tuning of some of these problems easier.
> That is, there may be obvious tuning that can take place btw the leaves
> at depths 1000 & 2000.
This is very true!
On April 23 Gunnar wrote:
> 2. I believe that the owl code would benefit from a major redesign,
> now that we have gained experience of this kind of reading, but that
> is not quickly done. A simpler experiment would be to run the owl code
> twice, where the first round has the restriction that only critical
> moves and eye-making patterns may be used. Groups which can live under
> this restriction would count as strong relative to ones which have to
> run. The main work here would be to go through the patterns and decide
> which ones have a definite eye-making purpose.
As for going through the patterns and deciding which ones have
definite eye-making purpose, one approach would be to assume
that they ALL do EXCEPT for those having to do with escape, to
start with the D13xx patterns. So I think the D13xx patterns
would be a good start on such a list. Let us call them escape
I am going to suggest a different reason for making such a
classification, independent of Gunnar's suggestion. I am going
to suggest another way in which the escape patterns should be
Escape is currently a problem with the owl code. On the
other hand, we need more escape patterns. On the other
hand if one starts adding escape patterns one quickly
runs into larger variation trees.
Let me give two examples to show that we need more escape
(1) nngs1:36 is a case where escape is overlooked, so the
W dragon at N10 is overlooked.
(2) owl:149 is a case where the test currently passes but
an obvious escape move is not considered so the variations
generated are totally bogus.
I would propose that a parameter is passed during reading
which counts the number of calls to escape patterns in
higher nodes of the tree. This parameter is incremented
when an escape pattern is. (This escape pattern count
would have to be added to the cached data.)
If this number exceeds a small constant, perhaps about
two, then calls to owl_determine_life are not made. Thus
once the dragon embarks on an escape it has to continue to
escape, rather than trying to make eyes. Allowing one or
two exceptions would help in situations where the dragon
can make an eye by threatening to escape.
- [gnugo-devel] Owl patterns,