Re: Shallow Or Deep Search?

["P.J.Leonard" <P.J.Leonard@xxxxxxxxxx>]

David Fotland wrote:
> 
> The tactical search is called from within the full board evaluation, so
> I never apply the full board eval at tactical terminals.  The tactician
> only uses a subset of the data structures, so the data is not even valid
> at the end of a tactical search if I wanted to do a full eval.
> 
> Have you done any tests of your life and death engine?  Mine is not very
> strong.  Gets all of graded vol1 and about 90% of vol II, when I let it
> have about 20-50 nodes per search.  This is mostly due to bugs in accurately
> evaluating eyes, especially when they have dead enemy stones in them.

 I have not done any testing. My move suggestion is far too simplistic
at
present (just try them all).
My detection of death is when a string dissappears. Life is detected
when the oppenent can make no more moves or a certain number of
liberties have
been acheived. This is of course not good enough it means that at best I
would
be taking an order of magnitude more search nodes than you. 
All this makes it a more interesting challenge for thesearch engine :-)

 I do detect and merge duplicate board states which reduces that size
dramatically. 

 Is there anything published on move ordering for life/death problems
also a what point
it is best to do a static analysis ? 



 cheers Paul.