Usually when something is not statistically relevant you can also prove it. I am not sure how you would do it in this case. It probably works to some limited e

Usually when something is not statistically relevant you can also prove it. I am not sure how you would do it in this case. It probably works to some limited extents: for example, I am sure they can guess rather accurately whether the candidate is male or female. So I do see a difference between a non valid tool and a tool whose efficiency is hard to evaluate. Now, don't get me wrong, I wouldn't use graphology.