Re: [gnugo-devel] reinstating value(100) owl defense moves?

[Arend Bayer]

On Wed, 1 Sep 2004, Evan Daniel wrote:

> Thinking about the owl code, it occurred to me that there could
> perhaps be significant gains to be had by reinstating and judicious
> use of the special case of value(100) for owl defense moves.  That is,
> if an owl defense pattern matches and has value(100), owl_defend
> declares victory without further reading.

> Clearly, bad patterns here could cause significant problems, so
> caution would be needed.

> |.????
> |.????
> |.O???
> |*.O??
> |..O..
> |.....
> +-----
> 
> :8,-,value(100)

That is already a good example for caution required:

|......
|.XX.X.
|.OX...
|O.OXX.
|..O..X
|......
+------

Black at 2-2 kills this.

> However, I believe the potential gains are
> also quite significant.  Currently, for example, owl reads out the
> life and death of the tripod group every time it appears in a game. 
> It gets it right, and considers all important variations, but it takes
> over 400 owl nodes to do it.  That means that there could well be
> cases in deep reading where owl would arrive at a knowbly alive tripod
> group, but be unable to read it correctly due to depth or node count
> limits.  A handful of patterns could at least partially fix this:

I am a bit unsure. My fear is that we would have to add some 100
patterns to see some 1% decrease in owl nodes overall in real life. Given
that it also adds to the time spent in matchpat.c(), and that it adds
complexity in the code.

Maybe you could try out how much your handful of patterns saves when
replaying "real life" games. Then we may have a guess how much saving
one could get with, say, some 50 patterns.

Arend