Re: Shallow Or Deep Search?

[Dave Dyer <ddyer@xxxxxxxxxx>]

I've used deep search almost exclusively, although I have experimented
with iterative deepening.   There are two big problems with shallow 
searches.

1) Results from searches that are not "deep enough" give results of 
such poor quality, that they are tantamount to worthless.  Only when
you read *some* definite answers (ie; "it's dead") do you get results 
that are reliable enough to usefully direct the next level of deepening.

2) the overhead in generating, regenerating, and evaluating positions
is substantial, so the cost of the shallow, indecisive searches with poor
cutoffs is very substantial.  This is also why searches based on alternatatives
to alpha-beta such as "proof numebr" search are problematic.  These other
search strategies invariable require the search engine to be agile about
switching fom one part of the tree to another.

--

The nemesis of search engines continues to be positions where a
solution at depth n is possible, but the problem solver is exhaustively
evaluating a different branch where the solution doesn't exist within
the absolute depth limit.