Wild speculation, which shall not provoke hints, but which people love to read anyway:
* Order of letters doesn't seem to matter
* Symmetry of individual letters seems to play a role.
* If you can mirror letters vertically (B, E), it tends to make koans white. Vertical mirroring means there is a horizontal axis of symmetry.
* Horizontal symmetry, or both symmetry axes present (W, V, H, X, O) don't contribute or maybe make actively black. If there is something to this hunch, then probably both axes present makes them cancel out each other.
Guess: AKHTBN iff count(vertical-symmetric letters = horizontal axis of symmetry) > count(horizontal-symmetric letters = vertical axis of symmetry). <-- Fails for WRAPPER, which is white.
WRAP black, WRAPP white. Also Y white.
New koans:
UBLE
BBELU
YO
YOO
-- Simon