> The basic idea is this: 90 degree rotation (to the right) is represented as
> a circular shift (to the right) by 1/4 of the key length.  mirroring the
> board (swap left and right) is done as reversing the order of the bits in
> the key.
>
> Distinct hash values around the board would have to share the same rules.
>
> Picking a somewhat arbitrary example (on 19x19), here's some candidate keys
> (kept simple for manual typing)
>
> A2   = 0x 01 02 03 04
> B19 = 0x 04 01 02 03
> T18  = 0x 03 04 01 02
> U1   = 0x 02 03 04 01
>
> T2   = 0x 20 c0 40 80
> B1   = 0x 80 20 c0 40
> A18 = 0x 40 80 20 c0
> U19 = 0x c0 40 80 20
>
> Points on lines of symmetry (such as C3 with 4 equivalent points or the
> unique tengen) need more care with how they're selected).

That's the same system I used in my first Go program, and it appears
to also be the same as what is in the paper that Remi linked.  I
didn't use it for full-board hashes, I used it for patterns.
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to