Re: Advice on evaluation

[Henrik Rydberg <rydberg@xxxxxxxxxxxxxx>]

Dear David,

The problem you address seems to me to be of a very general nature,
and a very tricky one indeed. I'm not sure if the word 'Hysteresis'
has been used for it, but that's what I think it is.

If the value of a position is symmetric with respect to who is to
move, there is (as you point out) in principle no need to split
the evaluation into a black/white pair, but you may save one ply.

On the other hand, if the posisiton is really *not* symmetric with
respect to who is to move, you really would need to split the
evaluation anyway, and you would not gain one ply, only get a
better evaluation.

Since I believe positions to be non-symmetric, i.e., contain
hysteresis effects, I wouldn't think that change would lead to
a one ply save. Hence, I would advice against such a change in
the existing program.

Boring as this is, I must add that I would have made the changes
anyway. :)

Seriously, this is a very interesting and deep topic, and I'm sure
someone out there has given it a lot of thought. Maybe among the
game theorists?

Kindest regards,
Henrik

-- 
Henrik Rydberg (http://fy.chalmers.se/~rydberg),
Department of Applied Physics, Chalmers University of Technology.