You can think of an example where things are classical, the particles start with some definite orientation randomly determined at the start and if they are within say 45 degrees of the angle of the detector they go one way, over 45 the other and it's not so hard to figure in that case it will vary linearly. As to how to prove the general case I don't know. Try Bell's paper?
As an aside I don't think the classical 'hidden variables' situation has a prefered orientation which contradicts the premise of the featured article that you get the odd entanglement effects so as to not have a prefered orientation.
