If it's just a human invented squiggle of some sort and doesn't relate to anything real then it's power to predict reality ought to be random at best. If, on the other hand, it does relate to something real, say some abstract principle then abstract principles must be "real" in some sense and the difficulty comes in defining in what sense they are real and how they interact with the concrete. It's an entirely open question, as far as I'm aware, with massive holes on both sides of the argument. The "it's a human invention" people currently have the upper hand in the philosophical debate but they definitely struggle to explain how something we just made up, which doesn't relate to anything that can be described as real, turns out to be useful.