One game idea almost everyone gets is to make a game about managing a stable of characters. Instead of actually doing whatever they would normally do, you abstract it and make the game about managing a group of them. Common examples are stables of sports players, fantasy heroes, or professionals.
There are some advantages to this approach - low content, easy abstraction, etc. But there are fundamental problems. Deep problems.
The first and most obvious is one of play complexity. Most games have a play complexity arising from actually doing the things these people do: running around performing in a complex environment. Abstracting that out leaves you with a serious lack of complexity - both in play and potential plot.
You hire people, you equip them, you send them out, they report back. It has no teeth. In order to give it any kind of interest, you'd need to have every mission have a huge, huge, huge, huge, HUGE variety of finicky detailed results, so that it matters whether you equip someone with a spear or a sword, or whether you tell them to be aggressive or friendly.
Then you would have to communicate these details to the players. Bleah!
There are a few solutions, but it's pretty damn rare for anyone to even notice there are problems in the first place.
One potential solution is that characters only tell you the finicky details they notice, so who you send actually determines which finicky details you'll get reports on. That's interesting, but it's still a very slow-paced game and there's a lot of algorithms to get to work.
Another potential solution is to add in a minigame for the missions. They're hunting dragons or playing basketball and you're playing, say, minesweeper. Depending on who you send armed with what, your minigame has slightly different statistics. Field dimensions, mine count, error tolerance, number of times allowed to check whether a tile is a bomb or not, etc.
A third solution might be to let the player make all the "top-level" choices on a mission. He can't control them in combat, but he can choose whether they enter combat or try to talk their way out of things. This has a few flaws, the worst of which is that it actually doesn't add much complexity unless you basically turn it into a minigame.
Anyhow, any way you run it, in order for there to be any emotional investment, the characters the player has in his stable have to have some kind of personality. This could be handled as a highly complex multi-axis algorithm that takes ten years to program, but a better solution might be to have, say, a dozen different personalities and a dozen different speech patterns, match them up at random, and add friendships, enemies, phobias, and preferences basically at random. Out of the garbage the players can synthesize a personality - it happens all the time in games with generative content.
Well... it might be a fun game, for all that. Once the faults are addressed. Hurm. Still some faults, but if we add the complexity back in, we might have something acceptably interesting...
What do you think?