When I'm in the early stages of a project-- or, to be more precise, when I've just finished a project-- I flail around a lot, falling in love with and then abandoning concepts left and right, until I put enough effort into one project or another that I feel like I better see it through. It's awfully annoying for those around me, not to mention those working with me on one or another of these projects, not to mention on me, period.
I've restarted Ultra Geist from the ground up about six times now-- a bad sign for my "quick little game"; I've started three or four completely different concepts for Yo Yo Games's Competition 5, one of which I documented earlier on this blog, four of which were shmups, and put a considerable amount of time into both coding and spriting all the various iterations. And they've all been pretty much tossed away.
My current plan is to take a couple of new ideas for Ultra Geist, combine them with some ideas my recent slew of abandoned shmups have stirred up, and basically make Ultra Geist the competition game I've been meaning to work on. We'll see how that works.
Right now, I'm trying to figure out the various types of non-boss units to populate Ultra Geist.
I definitely want a popcorn unit-- a term I stole from Warren Spector, an enemy that doesn't pose much of a threat to the player and can just be swatted away like a piece of popcorn. Think: a goomba: slow, dumb, easy-to-kill. Get a lot of them on a screen, and now they're more dangerous, but not overwhelmingly so.
Conversely, I want to have a sort of anti-popcorn unit. I'm not talking about a tank or damage sponge, though one of those might be nice, too-- I'm talking about a numerous, dumb, slow, and mostly harmless enemy that can't be killed easily, that can't be harmed directly but only through harming other enemies or causing other things to happen that then cause things to happen to the anti-popcorn unit.
This is because one thing I want to do is create units that interact with each other, a sort of web that the player has to untangle and then exploit. How well I can pull this off in a fast-paced shmup is the million dollar question, especially because I don't want it to devolve into a trial-and-error, figure out what-does-what sort of game. I want it to be intuitive, and I want there to be some strategic depth that's a step above "kill these things to do this".
I want the player to have options, different ways to approach it, different ways to maximize their play-time and different but related mechanisms for the players to exploit. The whole point of the game, again, is gaming the system, is looking for "bugs" and holes, is, well, "cheating".