What name do you use for an approach that probably gets a better than random success rate (without guaranteeing that)?

I’m leaving out the ‘pseudo-scientific’ part of GP’s description, since it doesn’t entail anything technically.