Okay, if you've never made a game before, I do have some advice that's always appropriate:

• be ready to work long and hard, because even a small game isn't easy to make;
• be ready to scale back your scope -- you won't be able to put in all you want, no matter what;
• it's better to make one 5-minute game that's complete and polished than several epic tech demos that never go anywhere.

Also, whatever you do, make a game you'd like to play yourself. Don't think of a generic audience, think of yourself. If you like the result, other players will, too. Besides, you're going to playtest your own game again and again, until you're sick and tired of it. If you don't even like it, how will you find the will to persist?

it is a good idea to search for some other people to help you with the game.

for example one is good in programming, another one can do good graphics, also planning the game is a part for itself.

i planned a 3d platformer game like supermario 3d land/ world with the BGE.

but before i start this, i begin with a smaller game, that does not need so much game logic (more game logic = more programming), graphics.

i do this, because i am a beginner with 3d games too.

another reason for this is, that the bge is not so powerful at the moment but it is getting better. so i can wait a bit.

with my smaller game i can learn a bit more about the BGE and how to use, i can set up a simple character to make a playable game out of it. 

later i want to concentrate on planning (which features to be added) and a bit on graphics (gimp, blender) and search help for other things like programming or sounds)

this helps if the game gets bigger and more complicated with a lot of features.