RE: computer-go: Dobbelt moves

[Petersen Kjeld-WFKP1396 <Kjeld.Petersen@xxxxxxxxxxxx>]

Hmmmm

How about calculating different values from the pairs that could be played ?
- Possible influence
- Possible liberties
- Possible stability (the ability to stay on the board)
- Possible territory
- Possible threats
- Possible flexibility
- Possible connectivity (what ever that is)
- Possible development of a stone or group
and so on

and then use fuzzi-logic to determine which kind of move we have !!!
- Sente
- Gote
- Eye making / stealing
- Life giving / taking
- attack / defence
- crap moves (defently bad moves)
or others

now with a knowledge of the status of groups on the board 
- living
- dead
- influence giving
- territory giving
- aji
we can choose which move would be best for a group or attack on a group.

My point is that the influece of a move should not only be calculatede by the influence the move it self gives, but as the possible influence a move will give ?

One of the main points in playing Go, as I see it, is to know the possibility of development !
So besides the normal board with the position of the black and white stone a program should hold a record of the posible developments of every stone and group on the board !

So if a move serve more than one purpose or stronger purpose, it is better than a move with less purpose.
This move is life giving to one of mine groups and it attacks a yours group, that will change the direction of development for your group against one of mine group with good influence/stability/living !

You can have a single move with (ex.) 3 minor purpose, that have a higher value than a single move with a 1 strong purpose.

By strong purpose I mean something like this:
.............
...ooooooo...
...oxxxxxo...
...ox.a.xo...
The move at (a) above is a move with a strong and defined purpose, while
................
.......b........
................
.......x........
................
................
the one-point-jump out at (b) has a more loose purpose, but if it serve more purpose like
- connectivity
- influence
- attacking
- flexibility
this move will have a higher value than the strong move !

I better stop now, otherwise I will go on and on for ever !!!
Sorry for the long article !!




-----Original Message-----
From: William Harold Newman [mailto:william.newman@xxxxxxxxxxxxxxxxx]
Sent: 27. februar 2001 14:02
To: computer-go@xxxxxxxxxxxxxxxxx
Subject: Re: computer-go: Dobbelt moves


On Tue, Feb 27, 2001 at 09:05:22AM +0100, Petersen Kjeld-WFKP1396 wrote:
> This is my first mail into the computer-go mailling list.
> So I want to contribute with a little teori of mine !
> I call it "dobbelt move testing".
> 
> If I should write a program that should check the importancy of a move,
> I would make a move follow by a move in the same color.
> 
> The value of the first move is equal to the influence of the sum of
> all possible or good moves followed by the first move.

> 
> Ex:
> 
> ...........7
> ...........6
> ...........5
> ....OX.....4
> ....OX.....3
> ...aOX.....2
> ..aax......1
> abcdefghijk
> 
> This is the best example that I could setup for the moment.
> 

> The value of the (black=X) e1 move is equal to the sum of the
> influence black making the moves
> VALUE_e1 = (e1 & d1) + (e1 & c1) + (e1 & d2)
>
> The program should now only make that first black move, and then
> test with 2 white moves to see have good the first black move is to
> resist any moves from white.
>
> Does any one understand my example ???
> 
> The dobbelt move should be make to test how important a sente move is.
> There could also be the possibility to test if the program should
> prevent a sente move (Gote) and test how important that move is !!!

The analysis and examples in _Mathematical Go_ (Berlekamp and Wolfe,
ISBN 1-56881-032-6) show that the importance of moves is only
partially ordered. Since real numbers are totally ordered, you will
get less-than-ideal results if you try to express the importance of a
move with a single real number. Trying your theory on the examples in
the book would probably demonstrate this.

-- 
William Harold Newman <william.newman@xxxxxxxxxxxxxxxxx>
software consultant
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C