> I have seen many PhD students, including myself, arrive with half-baked ideas that go nowhere.

Another way to look at it is that failure is a part of learning, and chasing those half-baked ideas can build experience (by finding out what doesn't work).