I've spent the last several days refactoring my OUYA game. It seems strange to spend so many hours working on the code and the great accomplishment is that it still works. I just felt the need to explain why I do this and how it works for me.
I believe that I am a very results-oriented person. Not to say that the ends always justify the means, but I recognized many years ago that getting a product working is much more important that making sure that all the T's are dotted and the I's are crossed. I've also been known to get forgiveness instead of permission on many occasions.
You can be writing the most elegant, perfect, optimized code in all the world but if no one ever sees or uses the product then what have you really accomplished? This is true for everything, not just software products. Once (1998) I decided to replace the stock 3.8L V6 engine in my 1989 Mercury Cougar with a 5.8L small-block V8. I had 3 months until I needed the car to run again so I could go back to college. As far as we could tell, this feat of automotive engineering had never before been accomplished. The engine did not physically fit in the car. But I wasn't building it to put on display, so we hacked and bent and cut and torched and pounded with a hammer. It wasn't pretty, but it ran. I drove it back to college and almost every day while I was there. I also spent the next 3 years fine-tuning it. Parts were replaced with new ones as they were fabricated. The more powerful engine totally blew out the rest of the drive-train in time, but that got fixed as well. I could have spent those 3 years making it perfect before it ever left the garage, but it was much better to be driving it all that time.
Basically, I learned that success means shipping a product. So how does this relate to hacking and refactoring software? It's pretty much the same thing.
I wanted to get stuff on the screen of the OUYA and get things moving around with the controller as quickly as I could. I needed to see it working, even if the code was less than perfect. I needed to see those results so that I could continue. Letting it "sit in the garage" for too long would just discourage me and I would lose interest.
However, I know better than to leave the code that way for too long. If I am going to build a complete game, my code has to be structured such that it can be built on. It can't be brittle. It has to use the proper design patterns. So every few weeks, I stop adding new features and refactor. I clean up the ugly code and make sure I have a solid foundation before building more. But in the mean time I also have something that my kids and I can play (You don't have to release your product to everyone to count it as a release. My Cougar only had one user, after all).
It's just what works for me. And it works for other people as well. Check out the works of Kent Beck and Martin Fowler. They've written lots of stuff on design patterns and refactoring code.
Also, read this paper.
And as for my Cougar, I no longer drive it but it taught me a lot about successful engineering projects.