Hopefully you've surfed up to a few web sites for a deeper look at the tools and technologies you can pull off the shelf. If you're smart you've figured out that you could spend a cool half-million dollars and most of a year on half-baked technology, with tools that work on every alternate Thursday, and wonder why you can't get more than 12fps.
There's another drawback to rolling your own engine. Publishers generally won't touch you with a ten foot pole if you tell them you are about to write your own engine. That is, unless your name is John Carmack. They want to hear that you've got one that you own or you've already licensed one. Publishers see that kind of situation as a high risk scenario. They are much more likely to require you to use some crappy engine that they had in their basement rather than allow you to write one yourself.
If you haven't been scared out of your wits yet, and are convinced you are going to write your own engine I certainly hope you've written one before and shipped a game. If you've done that you won't be starting over from scratch, which is a real plus. One piece of advice is to evaluate some of the middleware anyway, it will give you a good idea on how your own tools measure up. If they fall short, you can either step up to the plate and fix your old code or perhaps you'll just bail and buy something off the shelf. Either way, never assume you have the hottest thing going. Go find out.