yonax, you uncovered a nasty bug. I have fixed it and added the test cases discussed above to the kata.

This was due to an issue in my logic for solving the problem. It has been corrected, and the kata has been republished.

To confirm,

vector_affinity([None], []) == 0.0 => True

easy enough, but then shouldn't i also add:

test.assert_equals(vector_affinity([None], [None, None, None, None]), .25)
test.assert_equals(vector_affinity([None], [None, None, None, None, None]), .5)
test.assert_equals(vector_affinity([None], [None, None, None, None, None, None, None, None, None, None]), .10)

?

To me knowing that:

vector_affinity([None], [None]) == 1.0
# and
vector_affinity([None], []) == 0.0

should be enough to deduce that None is also a value that counts, and that [None]/[None, None] gets a .5 affinity.
again, it's easy enough to add the test case, but unless that test case addresses a specific flaw in the instructions or problem, i'm hesitant to do so.