Unless you care a lot about pixelling specifically, traditional media is a better option in that you can produce more, quicker. I personally have several clipboards with scrap A4 paper, and try to use exactly one piece of paper per drawing (so that I am drawing as big as possible).
My experience is that for building up skill, it's better to do many quick works than a few refined works -- I think of it in terms of 'the number of good new insights you can have per drawing' being limited. "Don't do anything knowingly wrong, but don't spend any significant time fixing a bad drawing either, just fix that flaw in the -next- drawing you do".
In terms of subject, I think that space is one of the most effective aspects to improve. This covers perspective, value, and construction. There are many possible approaches.
Some of them include:
* drawing cubes and rotating them in perspective without the help of explicit vanishing points. (see moatddtutorial's 5-part series on digital art basics, a demo of this is included IIRC)
* As above but with a more value-painting based approach (draw entire faces in their correct brightness, rather than the lines defining the cube)
* breaking down the object you see into 3d boxes, making sure that these are spatially possible (don't intersect other boxes in 3d space)
* There are also more curve-based approaches (eg. as shown in the first 20 pages of Vilpuu's Drawing Manual -- to oversimplify, "draw a wide range of blobs, connected or overlapping in different ways in space") which help to improve your sense of how to create volumes by conveying a sense of overlapping forms.
As I finish writing this and reread your post, it occurs to me that initially, you will not have a very good grasp of what you're bad at, so it's probably important to start out by doing a whole lot of different types of studies to locate where you are most lacking.