Re: [gnugo-devel] Patch: New move generator

[Inge Wallin]

Dan wrote
> I've been using strategy test 17 to examine Inge's
> patch. This is a good illustration where his code 
> works correctly. The issues in this example are
> with move valuation.
> 
> In a separate matter the moves at P5 and F6 are
> misevaluated as being worth some 62 and 54
> points. That's a separate matter. The issue is whether
> the point at N11 is found and evaluated correctly. It
> isn't necessarily the best point on the board, so
> as a regression the test is imperfect, but it's a very
> good example.

Yes, and in fact it's the one that I used when I designed the
feature.  

Now, frankly I don't understand the big problem with the patch.  If
Gunnar has different ideas of how to evaluate the territorial value
captured stones, then fine.  But my patch improves on the current
state of affairs and leaving it out only makes things worse than they
have to be.

I understand that the solution is not perfect and I have never
advocated against a better solution.  But no such solution is there
today and until there is, I propose that we use my patch.

> The move at N11 is indeed found by find_double_threats, 
> and the attack either move reason is added. Now, the
> issue is how the move valuation should be assigned.
> It appears to me that this move is worth about 20
> points.
> 
> After Inge's patch, we get:
> 
>     N11: 24.17 (followup) - threatens to capture N10
>     N11: 15.17 (followup) - threatens to capture K14
>   N11: 14.00 - attacks either K14 (14.00) or N10 (23.00)
>   N11: 24.17 - followup
>   N11: -3.55 - shape (shape values +2.00(2) -5.00(2), shape factor 0.864)
> Move generation values N11 to 22.53

There is one problem left: The 14 points awarded to the attack of K14
should be at least 18 since it saves some of our own stones as well.
I will refine the valuation to include this.

> This looks pretty good to me, except for some issues
> about followup (which we're not likely to get right
> anyway given the present paradigm).  

 I have continued the work on find_double_threats() and I have solved
the problem by simply removing the threats when I add the
ATTACK_EITHER move reason.  It works out pretty OK.

> I think Gunnar objects to Inge's changes in the
> valuation of worms, but in any case this example might
> be a good one for checking that the valuation is done
> correctly.

He does, but I don't understand how leaving it out would be better
than putting it in.

        -Inge