[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: "The value of sente"



On Wed, Jun 06, 2001 at 02:56:35PM -0400, Fant, Chris wrote:
> Could someone please define "the value of sente?"   Thanks.

Sure: 42

Seriously, it can be defined as the difference of the value of the locally
best reply and the largest non-local move, or something like that.

At least in situations where nobody is expected to read the game to its end,
one can assume that a move is 'sente' if the opponent has a choice of
answering somewher non-local, but even the best non-local reply (and
possibly the sequence that leads to) is worth less than making the (more or
less) obvious local reply to the move in question...

There are things confusing this definition. For example, in final endgame
where there are only a few isolated situations on the board, one can read
the combinations to the bitter end, and possibly play better than what would
be expected of mere 'sente' consideration... Or in the early game, when the
values of moves are so badly defined, that no reasonable comparision is
meaningful - then 'sente' can be said to be anything that the opponent is
(more or less) forced to answer... For example an approach to a corner
stone, answering it is very big, but occasionally even professionals play
elsewhere, allowing a double approach. The value of such a move is indeed
hard to set.


So, again, it turns out that go is not such a simple game as its rules may
suggest... Hooray for that!

-H
-- 
Heikki Levanto  LSD - Levanto Software Development   <heikki@xxxxxxxxxxxxxxxxx>