Re: [gnugo-devel] Patch: Value of edge josekis reduced

[Daniel Bump]

Inge wrote:a

> I don't think it has anything to do with J or j patterns.

No, definitely not. It happens in the middle game.

> This has to do with safety estimations of groups and especially how
> they interact, something that GNU Go is not very good at.  If a move
> connects two weak groups into one group instead and at the same time
> separates the enemy into two or more weak groups, then that is a good
> move.  Those things are not taken into account today.

Sometimes it has to do with safety, sometimes not. I suppose
we need a list of examples from games before we can really
discuss this. In many cases GNU Go makes moves implying an
intention of saving something, then lets it die because
something else lets bigger. Just to give one example,
move 98 in the game juka-gnugo-3.1.16-200112142153.sgf,
availailable from http://www.lysator.liu.se/~gunnar/gnugo/nngs/.
This is the kind of thing I'm thinking of.

> If we did even a small amount of work on this, it would make GNU Go
> much stronger immediately.

I think it might be more work than a small amount but I do
think this is important.

Dan