Just pick a game or three that you'd like to tackle, and go make them! Feel free to experiment with different tools, engines, SDKs, even languages - the more equipment you have to work with, the better your chances of being able to implement your designs down the road. In particular, don't feel confined to C++; if it seems like you're spending more time than you'd like wrestling with mundane stuff instead of getting games finished, then by all means swap to something else.

Other than that... just hack away, and pay attention to things that seem unpleasant or sticky - those are areas where you might be able to improve your programming. If you catch yourself feeling like "there's got to be a better way to do this" then heed that instinct - it'll serve you well. Don't worry too much about getting it perfect or even all that "right" to begin with; the best way to learn how to improve is to make mistakes and analyze them.

Final thought: keep it simple. Do the simplest thing that could possibly work. It isn't worth wasting time on complex reusable architectures or engines or whatever if that gets in the way of you doing what you're interested in.

I see this question quite a lot on game development forums.  I have decided to start a new video tutorial series that will address this concern by choosing a simple 2D single screen game (like pacman for instance) and show how to make it from scratch.<br /><br />I've called out the public to suggest which game I should make and shortly I will be choosing one from all the submissions and start to code.  Keep an eye on my website!