Re: [gnugo-devel] mixed patch

[Inge Wallin]

Gunnar wrote:
> I've put up a patch with mixed contents at
> http://www.lysator.liu.se/~gunnar/gnugo/patches/gunnar_1_12.2.gz
> 
> - new move reasons ATTACK_MOVE_GOOD_KO, ATTACK_MOVE_BAD_KO, and so on
> - move valuation distinguishes between attacks, defenses, owl attacks
>   and owl defenses with or without ko

I haven't looked at the actual patch yet, but doesn't this approach
lead to combinatorial explosion?  Each attack and defense move reason
is combined with a ko status for it giving very many move reasons
which have to be treated separately in the code.

Wouldn't a better approach be to have one attack move reason, and then
a parameter which tells wether it works right out or if there is a ko
involved?  This way it would also be easier to change our handling of
ko once we decide to handle more than just two kinds of simple ko.
Approach move kos and multi-step kos are just two examples.  We
wouldn't want a separate move reasons for attacking a string with a
two-step approach move ko, right?

I was thinking of doing something like this with threats, were the
parameter would be how many moves we were away from actually executing
what we are threatening.  These two parameters could be combined so
that we could, for instance, tell the move valuation that "this move
is an indirect threat to kill that dragon with a ko where he has to
make the first ko threat".

Now, to valuate that one is a completely different matter... :-)

        -Inge