If you read about small-scale game development, this is a common practice. Danc (http://www.lostgarden.com/) talks a lot about rapid prototyping. Since games have a clear but poorly-defined goal (they have to be "fun"), it's easy to make a limited product around a proposed method of finding that fun, and then either going along with it or throwing it out if it doesn't seem promising. Thus, you can (and probably will) start over several times before making a complete, polished game.