There is a poll going on at http://www.happypenguin.org/
This poll is completely unnecessary. Andre Lamothe wrote a book eons ago on how to write a game. Apparently people don't pay attention to the people who have made successful games to learn how successful games are created.
That said, you need marketing and a publisher. Not just a good game. About half the work involved is marketing and distribution of the product. The other half is the development of the actual game: Storyline, graphics, genre, the little extras (or, in the case of a game like FF7, NeverWinter Nights, other RPG goes here, little side quests - although in FF7, Emerald Weapon was insanely difficult to beat), the look-and-feel of the default controls, music, audible and ambient sounds, enemies, proper collision detection, the menu system (never underestimate the importance of a good menu system), etc. all work together to make a unified game. Then, once the game is made, beta tested, and ready for gold, in comes the box art, CD art, and website design and art specialists - these are the people who make the game good.
I got to that site through the wxWindows.org site from an e-mail someone sent me. I'm a programmer and have generally been disappointed that there are no games for Linux _worth playing_ that Open Source developers have made. The core problem lies in the fact that most of the Open Source developers like the idea of a game but have never really written one. The best people for developing a game are those who have done it before by themselves. None of this team junk. They went out by themselves and proved themselves to themselves and their friends.
Let me give an example of a lousy Open Source game and a good game (that happens to be closed source):
TuxKart. First off, the name is uncreative. Every Linux person knows that Tux is the penguin and that tacking the name "Kart" on the end is a lame rip-off of MarioKart. Basically, a poor attempt at a clone of the real MarioKart. It goes quickly downhill from here. The website design is abysmal. It looks like a 5 year old made it. It gets all technical and the author is quite whiny and apologetic in the download area, "The TuxKart download is currently just under 5Mb - sorry!" The graphics are ugly. The screenshots are ugly (which means the game is likely to be ugly). There is a controls page, which likely means there is no menu system in the game. The icons are ugly and pixellated. The background choice doesn't match anything in the game. Bascially, the website is completely lacking unification with the game.
Then, once you manage to swallow the fact that you have to compile the game (no pre-built binaries with a handy installer) AND install it yourself (writing a good GUI-based installer for a game is an art in itself), you are left with a shell and an empty feeling about the game...if it works. The website even states it might not work if you try the installation instructions. The website sets users up for failure.
TuxKart made the "Game of the Month" on the "Happy Penguin" website (Rule #6 of website design broken: Acronyms lose users, so I recommend losing the acronym "GotM" - users don't have a clue). This is the sad state of Open Source Linux games.
Here is how to make a good game and it would do game developers well to take heed:
The first thing you will notice is how professional the website looks. A nice change from the TuxKart website. Go back and forth between the pages a few times to see what I mean. For some of you, this is a foreign concept and one you had better get doggone right if you want to make a successful game.
Now there is one strange thing about this website - it coincides with the actual game. Graphics you find on the website are used in the game and vice versa. In fact, the game should bring you frequently to the website and the website should bring you back to the game - that is how unified the two must be to be successful.
UT2004 comes in a shiny box with a CD and high-gloss manual. Everything there is unified with the game and the website.
Do you see a theme here? I haven't said a word about how the code should be written and yet I'm a developer. See, I have the end-user in mind whereas 99% of all other software developers don't. What you must do to make a successful anything, whether it be a game, application, or tool, is make it so that the target audience feels comfortable about installing whatever it is. Keeping the look-and-feel consistent to what they are familiar with encourages them to install the game and give it a whirl. Irregardless of what the Open Source developers around here think, users WANT to just pop in the CD and have it practically install itself automatically. They don't want to 'make install' anything. In fact, they would prefer being clueless about how it works - they just want it to work. Most people playing games are playing just to either burn a little free time or to relieve stress. Know your audience. Games, especially, should ALWAYS have a binary installer (or compile behind the scenes if you insist on distributing source code).
Look at UT2004 or any of this year's commercial games, the related website, the marketing tactics, and the community in general. These are successful companies. The wise, discerning software/game developer will ignore my rants and go analyze how successful games are made and come to the same conclusion I've come to and shared freely here.