Re: computer-go: perfect play

["Vlad Dumitrescu" <vladdu@xxxxxxxxxxx>]

After a good night's sleep, I gathered enough self-confidence in order to
make here a few statements that I believe are true (even if I can't prove
anything). I hope they won't be too controversial - but on the other hand,
better to start an academic dispute than be ignored! :-)

* Black has a winning strategy
* Komi is irrelevant for perfect play
* Handicap is also irrelevant for the discussion.

Now before jumping at my throat, read the details:

* Black has a winning strategy
    This is only supported by common sense. Since having to move first is so
widely accepted to bring an advantage, I refuse to think everybody has
wrong!

* Komi is irrelevant for perfect play
    In perfect play, since there are no probabilities to balance, one either
wins, or not. Komi makes some of the winning positions (for black) to be
losing ones, but since black can choose the right sequence, it will only
have less lines of play at his/hers disposal. Or if the komi is too large,
black can never win...

* Handicap is also irrelevant for the discussion.
    In my opinion, handicap Go is another kind of beast than regular Go. As
with komi, the placing of handicap stones is just a convention - it doesn't
really belong to the "pure" game.
    Besides, the meaning of "one stone handicap" is (as discussed before)
not the same everywhere on the ranking scale: between two 20 kyu, it doesn't
mean much, while for two 9p it more or less decides the game. Also, between
2 and 3 stones there is a different gap as between 8 and 9 stones. So all in
all, it's not a reliable ranking measurement.
    Also, when comparing to chess, handicap creates confusion (to me it
does, anyway) since chess lacks that feature. Of course, white could get
several moves by itself in the beginning, but I think 4 or 5 are enough even
for me (with an ELO of at most 83, if not less!) to beat most people in the
world, and maybe even the perfact player!
------

>I am getting the feeling that theoreticall correct go will be quite a
>different game from the kind of go that is played between mere humans.
 I fully agree with that. Adding that interesting play will only take place
between two perfect players that don't know that they are perfect players
(neither themselves nor the opponent). Otherwise the game loses any kind of
appeal... :-)

/Vlad