RE: computer-go: Relativ moves

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

I think using well-known move sequences to assist reading is a good 
thing.  Many Faces does this, since
every move generation pattern suggests not just a move, but a move 
tree.  The details of the representation
of the sequence is not so important.

David

At 10:21 AM 5/18/2001 +0200, you wrote:
> With this a program could also recognize the start of a string and then 
> play out the continueation of the string (or strings) to se if the program 
> should try to find a better move or not.
>
> A up-right ladder have this relativ string
>     O X
>     |\ \
>   O X X-O
>   |\ \
> O X X-O
>
> O X O
>
>   O
> [0000 0001] x=0, y+1
> [0001 1001] x+1, y-1
> [0001 0000] x+1, y=0
> [1001 0001] x-1, y+1
>
> If the program recognize this string, then it know it is a up-right ladder,
> and will look for at ladder break in the up-right direction.
>
> But this should not only be restricted to ladders.
> When I play games on a normal board I often uses well known strings of 
> moves to se the development of the game. This should also be applied to a 
> program.
>
>
>
>
>
> -----Original Message-----
> From: Petersen Kjeld-WFKP1396
> [mailto:Kjeld_Petersen-WFKP1396@xxxxxxxxxxxxxxxxx]
> Sent: 18. maj 2001 09:28
> To: 'Computer Go'
> Subject: computer-go: Relativ moves
>
>
> I just came up with an idea. And I want to have your oppion.
>
> We have the current board stored in a array. 1D or 2D that doesn't matter.
>
> For the search tree we set up a string of moves,
> where each moves is stored as a relativ position
> instead of absolut position.
>
> 1 byte can be used to store relativ moves with a range of +/- 7 in x and y 
> direction.
> bit    7 654     3 210
>     +/-x dist +/-y dist (0=+ and 1=-)
>
> So 18 gives a relativ position of x+1 and y+2, from the last move.
>
> By seting up a relativ string of moves we can have prestored strings with 
> different values.
> a string like [18,45,230,47,193,56] could be a standard string which gives
> Black a increase in value of 26 and
> White an increase of 17, when it is put onto the current board.
>
> So this is a good string for Black.
>
> Instead of using single moves, a program could start with set of relativ 
> string moves
> and see if any of them gives a good result.
>
> This is like joseki, but for the hole board.
>
>
> Kjeld
David Fotland