This is subtle but I would say it is more difficult to predict than random. The colleges have internal policies, consistency checks, institutional priorities, and so on that lead to their decisions. But most of that is confidential or only vaguely discussed in public.
For practical purposes then you have to account for this uncertainty, but a truly random model isn’t quite right either. And that explains why, say, 3-6 well chosen reaches might be a better idea than 20 reaches picked off a generic ranking.
So if you picked your reaches with care and wrote your best applications you have done what you can.