I started writing down ideas like crazy for my new sandbox open world game. But it seemed like a mess, I don't know what genre game I should focus on: because I have everything from medieval villages to modern cities, from nukes to planet exploration. They don't really fit well together, and texture pack designers can't really get a medieval texture for a modern thing. If I should really focus on one genre, what should I focus on?

Why not create a new genre? You don't have to neatly fit into a particular box. The trick is to start small (I mean really small). Implement one idea/feature at a time. Figure out which ideas complement each other and find ways to connect them via gameplay. Time-travelling, for example, could be a game mechanic that ties your locations together (ex. the medieval city to the modern city). Iterative design is a powerful thing, for sure.

Depends on what you want for your game and what aspects you want npcs to manage. All settings slightly change plausibility of certain economic systems, production sytems and objects and tools but other than that they are mainly a matter of taste, what settings you prefer in stories for example?

I rather like Onigiri Flash's suggestions, but if you're going to focus on a single genre--and doing so might have the advantage of reducing the workload inherent in creating your game--I'd suggest selecting a genre that you find particularly appealing--something that you particularly enjoy.

Of course, you're particularly worried about sales of your game, you might perform some market research--but this may incur the risk of selecting a theme that you're not terribly interested in, and that you thus don't put your heart into. Additionally, while market research may direct you to a safe choice of theme, I'm not convinced that it's likely to direct you to much more than that.

Another thought, albeit one that may incur significantly more work, and perhaps thinking about user security: Your game is to be a sandbox game, you say--what about building in support for modding, and allowing your players to implement additional settings?

Perhaps it's worth asking: What is your sandbox game about? (This can be narrative, thematic, experiential, or mechanical--for example, Minecraft might be said to be about building, as well as exploration and survival.) Put another way: What experience do you want players to have in your game?

You can take stock of your team's capabilities and limitations and see if that helps guide you to anything in particular.

For example, say you're really not happy with the models that you've come up with so far for humanoids but you can do a really awesome fish. Then maybe the whole thing should be set underwater. (Would that create a whole new an aquarium sandbox genre of games?)

I also suggest asking yourself what is it that you want to get out of the process of making your game? Is this project about making money? Is it more of a hobby? Do you want to practice learning a particular skill? Are there particular game elements that have you excited to look at over others?

While its really tempting to create all your ideas and just shove it into one game, what it boils down to is how fun is it for the other players?

All those ideas can also be extremely overwhelming starting from scratch. What I would do is pick one theme and start out broad.

Pick a theme that interests you the most. If they all interest you the same, then this decision shouldn't take you more than 5 minutes to make.

When you have your theme, start with baby steps. You'll soon learn what interests you and what doesn't. Shift on your toes.

Theme: Space
Start with a basic planet (planet exploration) where you have a character who can move around a small map and pick up items.
Then create a simple attack-death system so you can fight other players (or NPCs).

I'm assuming you want a multiplayer game, begin programming a SIMPLE multiplayer functionality (or try a library of some kind). This in itself is it's own beast, and will quickly test your dedication.

If you get past these steps, find around 5 players to test your game.
Slowly add/remove/change small features until you are building something both you and the other players find fun and enjoyable.

What language are you using?

what game world/universe do you want to build / do you want to play in?

that's whats you'll do best at.

if its just for fun, that's probably what you should make.

if its for profit, unless its something so out there nobody else would want to play it, or its already been done and you can't do it better, its probably what you should make as well.

If I should really focus on one genre, what should I focus on?

That's a bit like saying, "I like food, what should I buy?" ;) There are many options. It would be a good to narrow down the focus, both from a project management point of view and from the point of view of playability. Maybe the best thing is to ask yourself what game you'd like to play yourself? And make that.