A sample test would be nice, even though the kata is well explained.

Needs random tests

Random tests?

Updated the description

The a should have been :a.

It is now fixed.

Fixed.

In the first sample test:

Test.assert_equals(:a.to_ary, [a])

a should :a

You will need to optimise your code. For example, what is the big O lookup time for a C# List? Are there faster implementations?

It should be made clear that the order must be different every time the function is called.

I

Yeah I think you're right actually, that looks better, I shall rework this kata

The second example of your description shows that positiv angles have to be rotated to the left (=> neg. to the right - a point for your description?!). So the example for 45 degree looks a little bit strange to me (even for a rotation to the right - "swapped values?"):

[[1, 2, 3, 4],
 [5, 6, 7, 8],
 [9, 10, 11, 12]]

normally should be

[[4],
 [3,8],
 [2,7,12],
 [1,6,11],
 [5,10],
 [9]]

Or where i'm wrong? (i know what you are doing here - but is it really a rotation?)

variable "a" isn't described before calling it in the test cases

I struggled with this too. It's extremely unclear in the description how the fitness function works. However using
fitness.call(chromosome.to_i(2)) returns a rational answer when chromosome is a binary string.