Start thinking about your game with The 3 Questions (what's it about, what do the characters do, what do the players do). Pay attention to Vincent's comment, here.
In this interview, Vincent Baker was asked what the most important thing is a game needs to do in order to be successful:
"It needs to express your passion. Write about something that moves you, and I mean really moves you. Designing games is like writing fiction or poetry: design from your worst, meanest, ugliest, most hurt place. Or your most in-love place. Or your sweetest, most naive place. Design to expose yourself.
"Human contact, right? It should happen between the players, and it should happen between the players and the designer."
Ralph (Universalis) Mazza wrote this:
"The essential element of design that tends to get skipped is the part of the process where the designer makes hard decisions about what the game needs and what it doesn't. Assuming the game needs everything is just lazy design. Some games (like Multiverser) have a good reason for needing a wide range of things. Some games (like Trollbabe) don't.
"Knowing the whys and the wherefores of a proposed game design is the very very first thing the designer should do...LONG LONG LONG before worrying about the relationship between skills and attributes."
This post also makes some interesting points.
Ralph also wrote this: "Including rules for things you need is good design. NOT including rules for things you don't need is ALSO good design. Your design goals are there to tell you which is which. " He's talking about plausibility versus realism, in this thread.
Ron Edwards continues his 'through the door' metaphor of RPG design here. With a coherent design philosophy, "explaining role-playing or this whatever-it-is to non-gamers becomes easy to you, and intriguing to them - "hey, can I try it?" is their instant response."
"Have mechanics that focus play on what the game is about. Then gloss the rest." So says Mike Holmes in this post. Kind of a companion piece to Ralph's comment above. Mike's reply to failrate further down is also worth reading.
In this thread, Ben Lehman talks about basic design principles he's come up with:
1) Every act of resolution must have meaningful stakes.
2) Understand who can say what.
3) Know what your game is for.
.... 3a) If "realism" isn't what your game is for, why do you care about it?
4) Trust the players
5) Don't be afraid to innovate. Don't feel you have to innovate.
6) Write for you. If you don't love your game, every last scrap of it, by the time you are done, you have failed.
"I think it's very important to spend most of my time designing the setting first. I think all other aspects of the system should be integrated in tightly with the setting. If I don't truly understand my setting, how can I design an action check system that captures it's flavor?
When I've fleshed out my setting, I'll go through each design element and ask myself how I'm going to accomplish it while integrating it in with the setting. Each of these questions usually end up a page under the appropriate setting in my binder. " Comments from Roy in Structured Game Design.
Jared Sorenson has Three Big Questions:
1) What is your game about?
2) How does it go about that?
3) What does your game reward/discourage/encourage?
Lumpley's checklist of what to look for in the cool new games.
A resolution system to nick from Vincent, for Power!
Check out Tony Irwin's comment in this thread (he's #16). It's a great point about explicitly saying where the fun is in your game; about using the rules to spell-it-out.
What Vincent thinks the complete game should have: a checklist (including a focus on either violence, sex, children, money, God, or art).