Using Normals for backface-culling...

BTW.:
If I Do this, instead, it works perfectly...

Function drawSphere()
For i% = 0 To 15
For j% = 0 To 7
x0# = +rx#(i%, j%) * 100.00 + 512.00
y0# = -ry#(i%, j%) * 100.00 + 384.00
x1# = +rx#((i% + 1) Mod 16, j%) * 100.00 + 512.00
y1# = -ry#((i% + 1) Mod 16, j%) * 100.00 + 384.00
x2# = +rx#((i% + 1) Mod 16, (j% + 1) Mod 9) * 100.00 + 512.00
y2# = -ry#((i% + 1) Mod 16, (j% + 1) Mod 9) * 100.00 + 384.00
x3# = +rx#(i%, (j% + 1) Mod 9) * 100.00 + 512.00
y3# = -ry#(i%, (j% + 1) Mod 9) * 100.00 + 384.00
If Abs(x2# - x3#) < 0.01 Then
dx1# = x2# - x0#
dy1# = y2# - y0#
dx2# = x2# - x1#
dy2# = y2# - y1#
Else
dx1# = x3# - x1#
dy1# = y3# - y1#
dx2# = x3# - x2#
dy2# = y3# - y2#
EndIf
bf# = (dx1# * (dy2# - dy1#)) - ((dx2# - dx1#) * dy1#)
If bf# > 0.00 Then
;If rnz#(i%, j%) > 0.00 Then
If (j And 1) Then
Color 255, 0, 0
Else
Color 0, 255, 0
EndIf
Line x0#, y0#, x1#, y1#
Line x1#, y1#, x2#, y2#
Line x2#, y2#, x3#, y3#
Line x3#, y3#, x0#, y0#
EndIf
Next
Next
End Function

Why is that?
I hope it is just some programming error of mine, as opposed to a flaw in the algorythm itself.
(In other words, I am dtrying to use the rotated normals for lighting the polygons properly, as well as performing backface-culling.
But, as far as I can tell, and if there is no "programming error" or typo in the code, I must assume that the rotated normals are not accurate enough for performing back-face culling.
If that is the case, what kind of algorythm should I employe?)
Thanks a lot,

Rui Barbosa Jr.