I have to agree with Ashaman73. Think small goals. Feel good when you accomplish them. Try to think as little as possible about your finished game as far as mechanics go. Usually this leads me down the path of "Oh, I want this to work like this, which means I'll need this object, which will need to communicate with such, and oh I should make this interface..." etc ad infinitum until you're thinking about 100 different things that need done, and you shutdown from overload. I prefer to think along the lines of "Ok, where am I now? What is one thing I can easily add? or, What is the next logical thing to implement?" You add that thing, test it out, refactor, and boom, you just successfully got a little closer to your end goal, and you feel great about it. Sometimes I just have no idea what to work on, so I go back and look at some existing code and ask "Is this as clean, or as efficient as it could be?" and then sit down to clean up that one piece of code, which also improves your program and your ability to look at it and determine next steps.

In a nutshell, think baby steps, and put the blinders on for every step thats past the one that is next to work on. Start from the ground, and build up to the final game, rather than looking at the final game and trying to break it down to it's tiny pieces.