In a fit of nostalgia (I miss living in Japan) I bought Bushido the other day. I haven't had time to read it yet (although I did abuse my office printing access by printing it all out and treasury-tagging it into booklet form). But I have had time to ponder one of the great imponderables: How do you run a samurai sandbox game?
Other people more intelligent and verbose than me have written before about what you might call the "Superman Sandbox" problem. This goes, roughly, as follows: Superheroes are by definition reactive. They stop crime. This means that superhero games aren't a good fit for player-driven, sandbox play - at least, not readily. It's difficult for players to be the engine of the game when they are mostly just waiting around for somebody bad to do something horrible so they can stop it. And it is difficult to prevent that sort of game turning into a big fat exercise in teat-sucking in which the GM is just dollopping out missions for the players to take on. What does Superman do in the downtime while he is waiting for Lex Luthor to come up with his next hare-brained scheme? He mopes around Lois Lane and rings his parents. Great session, guys. Let's fast- forward to the good bit.
You can, of course, run a superhero game in which the players are not just waiting around for the bad guy to do something bad, and in which they use their powers towards ends that are not necessarily all about stopping bad guys... But then, you're not really playing a superhero game anymore, are you? You are betraying the genre.
Samurai games have a similar sort of same problem, to a lesser degree. Of course, if your players are going to be yakuza types, or perhaps ronin, they are free to do whatever they want. But if you want to do things properly and run a game in which the players are samurai, this means they have to be the retainers of some lord or other, which makes it difficult for them to be free to pursue their own goals in a way that makes sense "within the fiction". They have to wait around to be ordered to do things. Or you have to betray the genre in some way and allow them a level of freedom they should not really have.
You might argue, and it is a fair argument, that the samurai game does not require the lord to be sending his retainers on missions all the time. It doesn't have to be so GM-centric. The lord might just give the PCs the responsibility for doing certain things, within the scope of which the players have all the freedom they want. "Keep the peace in Tosa province" is an ambiguous mission statement which the players can interpret in all manner of different ways, given a varied set of conflicts which the GM has dreamed up going on within the province.
But at the very least this still requires you to, as Zak puts it, design conflicts. You can't just do the ordinary kind of sandbox setup you would use for a game of D&D or Cyberpunk 2020 or whatever, drawing maps and creating lists of NPCs and interesting locales and then just set the players loose in it. Why? In D&D, that kind of setup works because you know the PCs have absolute liberty to do what they want, and you as the DM only have to react to them and ensure your setting reacts appropriately. In our "Keep the peace in Tosa province" scenario, you would be required to add an extra layer - the potential threats to the peace. This is hard work, because you don't have any idea how the players are going to deal with those threats, or in what order. They aren't just doing the rather predictable things that rogueish PCs will do in a game of D&D (delve into dungeons, go hex crawling, etc.). They are thinking of complex ways to resolve complex problems. And this puts a lot of pressure on the GM to be able to improvise and maintain a high level of preparation and thought, session after session.
To put it more simply, once you've done the setup for a D&D sandbox game, each session more or less writes itself. The players are doing things off their own bat and the DM is reacting. But in a samurai sandbox game, each session is something of a mystery (or, let's say, what is going to happen in the next session but one is a mystery). You, the DM, are creating things the players are reacting to, and they have many more potential reactions than you can really anticipate. To put it even more simply and reduce it ad absurdum, it is the difference between:
A: There is a dungeon. The players explore it. (D&D)
B: There are a group of bandits in the South-Western tip of Tosa province. The players...try to kill them? Arrest them? Go looking for more help? Decide to spy on them? Agree to turn a blind eye in return for a bribe? (Bushido)
One is relatively simple for the DM to deal with; the other is relatively complex.