it was about half a joke. statistical evidence doesn't really exist for the type of problem since polynomialy computable numbers are countably infinite so you can't define a uniform distribution over then