PREAMBLE YOU CAN SKIP:
Sometimes people are confused about how a sandbox campaign works. Particularly on the issue of improvisation v. preparation v. randomness.
Obviously improvisation is necessary to run a sandbox, since PCs need to be able to go anywhere their PCs could go if they were really there and do anything, and players (at least mine) are not boring and so will think of things to do and people to talk to and things to hit with flaming oil that are:
-implied by the gameworld, but
-not prepped yet.
On the other hand, if it's all improvised or random and nothing is "fixed", then the possibility exists that PCs don't really have any control of their destiny since any choice they make is basically saying "Make something up now, Zak"--or at least they may feel that way--which is just as a bad. (A good all-improv DM can, of course, avoid this, by always making up something related to the PCs input, but that's a different kinda game. We're talking about a certain kind of game: total character freedom of choice, strict rules for player input. It's a lateral-thinking and problem-solving challenge rather than a creative writing or drama-school-improv challenge. )
POINT IS: when something big goes down, players should feel the way Humbert Humbert does when Charlotte Haze got hit by that car...
Within the intricacies of the pattern (hurrying housewife, slippery pavement, a pest of a dog, steep grade, big car, baboon at its wheel), I could dimly distinguish my own vile contribution...
The key to the balancing act is information. No sandbox is wholly sandboxy: players are being DMed because--though they want freedom of choice--they also want to encounter the unexpected; and the unexpected is usually unanticipatable and thus, in its finest details, unchoosable. The doctrine of choice in a sandbox doesn't extend to "do you want to walk into a dungeon and have there be a bugbear behind the left-hand door or the right had door?" and sandbox players wouldn't want it to.
However: a sandbox set-up is sandboxy precisely to the degree that PCs could have anticipated the appearance of a setting or story element through use of information available in the setting before the thing shows up. (Whether they did anticipate or not is a whole other thing. Some people would rather just open a door then listen at it. That's cool.) (And, of course, how sandboxy it is depends on how it's GMed, too, but this is just about the set-up.)
Note that "anticipation" also includes: if the players know that the next thing to happen is going to be the product of a certain kind of random chart (the same way I know that talking to a random person in a bar at 3 am is going to elicit different information from talking to a random person waiting in line at the DMV) this counts as "anticipation". Point is: I have information that can inform my decision to do or not do a thing, or to do it in a certain way.
Anyway, here's an example ripped from last weekend. This is my session, with an eye toward how much was prepped and in what way:
Session begins. I tell Mandy and her sister they are on the Isle of Oth (since that's where the group left off).
So this part of the sandbox is, on paper, from the GM's POV:
24 target NPCs/monsters with stats, personalities and interconnections ("this guy hates that guy" etc.)
A map with the location of all these peeps' domiciles and a ton of other buildings
A handful of partial maps of a few of these buildings
(More info on Oth here, though you don't need it to understand this example)
The players have a map. The players' map shows where the 24 are known to hang out, and where every single interesting building in the city is. This level of detail is unusual and unnecessary, but I used a shortcut.
Note on PC choice: I could've just let the PCs figure out where the 24 targets were, but then the game would've been about that, rather than about tactical decisions, which is how I wanted it set up. That particular set-up idea was all DM fiat--basically: if you want the given carrot (the 4-fold Crown of Oth worth 3000 xp) the most obvious choice is to play the tactical kinda game in Oth. If you don't want to roll this way, you still have choices, since just setting that game up required so many moving parts, which the PCs can (and have) fucked around with in various other ways. Or, to put it in gamer jargon: Oth is a location, and the attempt to get the crown is an optional plot threading through that location.
The PCs also know some NPCs who can tell them the "word on the street" about any of the 24 targets, though this is often vague, like: "The Lovers are very dangerous". To the degree it's vague and not part of a careful structure of which-NPCs-in-the-city-know-which-rumors-and-stories-about-which-targets it's less sandboxy (less information) than it could be had I put in more work, however this is offset by the fact that almost all of the entities can be met (or spied upon) by the PCs without forcing a confrontation. That is--PCs can gather a great deal of intelligence on most of the targets just by meeting them on their own, and the nature of this meeting won't force them to fight with them or even bargain with them. That's all pretty sandboxy.
PCs also know Oth is a city and behaves like one (dangerous at night, most commercial goods readily available, not all NPCs know each other, etc.) so any other plausible objective drawable from that is available.
Since only a handful of the homes of the 24 likely targets on Oth are mapped--and those only partially, the rest would be a combination of improvised-based-on-what-I-know-about-the-inhabitants (like it's a crow-demon's house so...) and generated using random city tables and whatnot. For example: the PCs have no way of knowing what's in the crow-demon's house (unless we're assuming the interior is mundane enough to match the exterior, then they know a little), but they do have enough information to know it's a crow-demon's house, and that implies a thing or two. Maybe enough to make a decision.
Obviously it'd be better from an ability-of-PCs-to-get-info-before-it's-urgent point of view for me to have all these locations fully mapped but, hey, I got a life here. (PC freedom is hard work.) So far the PCs haven't eaten through the "wall of information" I've already built so we're all good.
So much for Oth. Mandy has been exploring the Isle with a lot of other people (not her sister) and has more than one PC, so when the session starts, she decides to leave the island and go somewhere else.
But where? They have a map representing roughly what the players know (and they are neither cartographers nor philosophers, so their knowledge is vague and the maps are vague) about the land west of Oth and another one for what's east of Oth (the one at the top of this post). If they ask for specific distances between familiar places in days or miles, I can tell them.
The west of Oth map has a lot of places they've been or know about--the whole campaign so far is on it. The east of Oth one is vaguer, but they could ask a sailor or trader for some info and find out more than they've ever bothered to--like: there's a chain of old island fortresses, some are still inhabited, the Realm of the Negatsar lies far to the northeast, stuff like that
I don't know a lot, but I know more than them. If I know, say, 10 things about what's going on over there, they could find out about 7 of them. Basically: the subgenre of adventure that happens over there, in different directions. My idea is: they'll be at sea for long enough that I can decide what's going on over there more specifically before they get there.
Because of my own limited information (and prep) and my calculation that this level of prep roughly matches what somebody like the standard Vornheim-born PC would know about this remote land anyway, right now their east-of-Oth decisions are limited to things like "Do I want to go to pseudo-Hungary or not?""Do I want to take a long sea journey to 1001 Nightsville or not?" There are no distinctive plot hooks or distinct objectives pointing in this direction, but then that's never stopped my PCs before so I don't feel too bad about it yet...
But they're not going that way, they're going west.
Mandy's sister's never been on Oth--she has a PC who rolls on the Flakey PC table and discovers she's been offered a military command during the weeks she's been away.
She decides to sail back across the sea to Vornheim and help out with the skeleton invasion going on back there. (Available information suggests the skeleton invasion is probably going to go ahead and fuck Vornheim right up if the PCs don't intervene. Which is ok with most of the PCs. This invasion was, incidentally, triggered by PC action, and the PCs were warned--if a bit vaguely.)
So Mandy and her sister sail back toward Vornheim. The sea is, to some degree, a known quantity. Mandy got to Oth in the first place by spending 6 days at the mercy of the Wavecrawl Kit so knows the drill: Each day you roll, weather can happen, sea monsters can happen, other ships can happen. Other ships appear on the horizon and can usually be investigated or avoided.
So they ail back and have some random adventures at sea. Near the end of the session they meet a (randomly rolled) ship full of refugees and pick up a (randomly rolled) passenger and the PCs decide to change course and head for the Goblin Empire. (I need to write a post about Girls In A Sandbox which will be about how girls change their minds a lot.)
Assuming the destination they chose, the PCs have 3 options:
-go by sea for 4 more days,
-land at a port that was dangerous last time they were there and go overland for 4 days,
-find another port, which would take more than 4 days and might force them to find more food, then make landfall, then walk overland.
They decide to go by sea. The session ends.
Well I know they'll start in the sea, so I've got 4 days of rolling on the Wavecrawl Kit before they hit land and I have to start figuring out how the goblins react to them. So far I've got 2 locations in Goblinland prepared plus I know generally how things in the Goblin Empire usually work politically, socially, biologically, geographically, magically and architecturally. These 2 locations can go anywhere I want UNTIL such time as the PCs start getting info about them, at which point they get nailed down to a specific place.
If the PCs sail somewhere in the empire I don't expect, then I'd roll here. If they manage to sail, land secretly, move inland, pass my prepared locations, and move toward an inhabited (i.e. "non-random-wilderness-encounter-having" location) all before the session ends then I gotta start making stuff up during the session.
That usually works out fine but it just means the PCs have no way of reconnoitering what's going down in this place before they get there (at least until I know they aren't pointed toward some place I know and start making shit up and they start asking around so I can feed them parts of what I made up) and so they have fewer opportunities for making meaningful choices about the tactics and tenor of their adventure at that point.