It all breaks down to understanding the economics of your game; a more abstract economics. As said in other posts, you really have to understand what want the game to value, reward, and punish. Only then can you really make mechanics that can properly fit.
You can look at games like DnD and gather by it's system that it is very focused on rigid combat. The bulk of the rules are there for combat, and the economics are very well defined for that. You take a look at Storyteller (White Wolf) and combat is treated as equally as any other skill via the mechanics. This changes what is valued in the game and what players will unconsciously (sometimes consciously) strive for. You can usually tell what kind of style a game is going for by looking at the mechanics. It's similar to going to a marketplace and seeing a bunch of a different kinds of Widgets and over half the vendor stalls dealing in Widgets in comparison to the many other things out there; you can safely assume that this town the marketplace is in takes their Widgets very seriously.
The main suggestion when dealing in mechanics and making rules is to look at the raw numbers, resulting statistics, and true economics. I've seen many people get hung up on the different dice rolling, strange abstractions of the statistics, and round-a-bout methods. There's very little point in trying something obscure and awkward mechanically to try to set your mechanics apart from the rest when the raw figures aren't that different in the end. This is especially true if it just makes things more complicated and difficult.