value of sente; life and death

[David Mechner <mechner@xxxxxxxxxxx>]

Jeremy Thorpe wrote:
> you're right about maneuvering, and it's another interesting topic.
> however, i wouldn't even have brought it up, but your approximation
> isn't even an approximation at all--here's why:  the average value 
> of sente in the first 100 moves is like 5.5 to 8.0 points (the 
> value of the 'fair' komi).  [...] anyways, it's no big deal, i'm 
> just saying.

Agreed it's not big deal, but you're wrong :) 

Who cares what the "average" value of sente over 100 moves is? The
variance is huge. If we each get sente a bunch of times when it's worth
5 points, but I also get sente once when it's worth 50 points (say,
because you left a group hanging), I'll win. The fact that the "average
value of sente" was 8 points isn't going to help you. 

There's a variant of go where white gets to play twice in a row, once
during the game, whenever he wants, and black takes a handicap to offset
that advantage. The handicap black needs to make it a fair game is
probably around 6 stones. In many, many positions, the value of the next
move is huge. 

If you drew a graph of the average value of sente over the course of the
game, it would start low (5-8 points), go up to a big hump during the
middle game where often the fate of entire groups depends on who has
sente, then trail off gradually in the endgame. 


> > Unfortunately knowing that a group is alive with probability 
> > .75 is not very useful for deciding how to play in a particular
> > instance. 
>
> but i disagree strongly.  always, no matter the strength 
> of the player, there comes a time when life is uncertain. 

I don't think we really disagree here. I'm just saying that in most
normal situations, the uncertainty you have of any sizable group's
status should be very, very low. In a normal game, if there's a sizable
group that you figure has even a one in ten chance of dying, you should
probably protect it. If you have multiple groups in serious danger, a
number (.5 vs. .7 alive) may be helpful for deciding which to protect,
but that's a very very bad situation to be in. 



> > Life and death is the critical subproblem of computer go.
  
 [...]
> but i see that your real argument is about the important 
> 'next step' where the computer finds the fine line that 
> we humans learned when we passed 20k, which is that there's 
> some times that some stones are simply going to die, and 
> there's nothing to do about it.

I think programs are already well past the 20k life & death level. I'm
talking about solving general middle game life and death problems of
normal complexity at the low-kyu level. 


-David

-- 
David A. Mechner            Center for Neural Science
mechner@xxxxxxxxxxxxxxxxx         4 Washington Place, New York, NY 10003
212.998.3580                http://cns.nyu.edu/~mechner/