Gerard 't Hooft has written an excellent set of pages on How to become a GOOD Theoretical Physicist - essentially his sketch of what a person ought to know before they can actually do any relevant work in theoretical physics. Yes, you might have a good idea, but unless you actually know what (say) a Lagrangian is it is pretty likely that you will not actually answer any real question. And in order to get to the Lagrangians you need certain math and mechanics.
I think the site needs a clearer dependency graph, but it is quite neat. Also, we need similar sites for other disciplines.