The Designer's Notebook: Sandbox Storytelling

Back in 1995 I gave a lecture at the Computer Game Developers' Conference in which I identified several problems with interactive storytelling. I reprised those ideas a few years later in a Designer's Notebook column called Three Problems for Interactive Storytellers. At the end of both the lecture and the column, I suggested that instead of trying to tell stories, we should build worlds in which stories can happen -- worlds in which players live a story of their own creation. The industry didn't have a term for it at the time, but what I was proposing was sandbox storytelling.

In sandbox storytelling, the idea is to give the player a big open world populated with opportunities for interesting interactions. The player isn't constrained to a rail-like linear plot, but can interact with the world in any order that he chooses. If the world is constructed correctly, a story-like experience should emerge.

Not everybody thinks sandbox storytelling is a good idea. The year after I gave my lecture, Bob Bates gave his own lecture at the 1996 CGDC called "The Responsibility of the Author."

One of the things he said was, "[Open-ended environments] may be fun to explore, but they do not fulfill the obligations of a story. There is no beginning, middle, or end. There is no pathos, no human drama, no greater truth to be gleaned from the hard-fought battles that the characters wage."

Bob recommended that we use a linear series of open environments instead -- what we now call a multilinear or foldback story, in which the player is compelled to go through certain choke points in the plot line.

However, Bob was assuming that in an open-world environment the player would have to go find the plot, and all she would get is a disconnected series of events. I think Bob was expecting that the plot events would be tied to specific locations, and if the player could experience them in any order, they would have to be unrelated to each other.

I'm not surprised that he made that assumption, especially back then. We're very used to mapping plots onto physical locations -- so much so that it's our default approach, and any other system is unusual. From Zork to Half-Life to Fallout 3, movement through space equals movement through the story. But to do sandbox storytelling, we have to get rid of this notion and think instead about how to create a plot that advances -- and maintains its continuity -- by other means.

The Grand Theft Auto games famously include sandbox play, but they don't do sandbox storytelling. Instead, you get the usual linear chain of missions; complete one and you get another one, and so on. It just so happens that the missions take place in a large open world, and you can abandon the mission and just wander around wreaking mayhem (or driving a taxi) if you want to.

Grand Theft Auto IV

In a way, this was what Bob meant by a linear series of open environments, except that instead of a series of different environments, the Grand Theft Auto games just give you new missions in the same environment -- although you do unlock new areas from time to time.

The Sims offers sandbox storytelling after a fashion. It gives you a world with a lot of stuff in it, and simulated people with varying personalities. As the player, you can make them interact and generate a (somewhat) story-like experience. Because the Sims don't speak English, most of the storytelling goes on in your head, but that's all right. You can make your own machinima, caption or record voiceover for it, and upload it to YouTube.

But The Sims uses a multipresent interaction model in which you don't have a particular avatar within the game world. To get a story out of The Sims, you have to manipulate more than one of the characters, rather than role-playing a single character. This makes you more of a creator than a participant. That isn't the way most storytelling games work, and I don't think it's what most people want from a storytelling game.

Computer role-playing games give the player a big open world, but rather than providing a single story, the world is full of quests -- essentially, disconnected subplots. I love Western RPGs, but they don't have quite the same feeling as a story with one plot. They're more like the legends of Hercules, or any other ancient hero who appears in several unrelated stories.

So how do we make an open-world game in which the player can roam around, yet still feels as if he's taking part in a story? First, as I said, we have to abandon the idea that the player will experience the plot entirely through exploration.

At the same time, traveling still has to be an integral part of the story; otherwise the travel will just be tiresome. Movies usually cut out travel time -- somebody comes out of their house in the morning, gets in their car, and in the next shot they're walking into their place of work -- unless the movie is actually about travel, as in a chase movie.

In the typical adventure or role-playing game, all the plot events are player-dependent; they don't happen until the player finds them and makes them happen. By using constrained environments, we can make sure that the player finds them in the right order. The problem with a plot consisting entirely of player-dependent events, as I explained in the original lecture, is that it feels mechanistic: the whole world just sits around waiting for the player to do something.

If you make the plot entirely player-independent -- that is, it goes forward no matter what the player does, even if he does nothing at all -- then the player tends to lose the game a lot. He's not where he belongs, or he hasn't done what he needs to do, when the dramatic climax occurs.

The trick in sandbox storytelling is to build the plot with a combination of player-dependent and player-independent events.

Keep things flowing no matter what the player does so the world doesn't seem static, but don't make it flow so fast that the player gets behind and loses the game (unless the plot is about finding a time bomb). Put a moderate degree of pressure on the player to act, but reduce the pressure if the player is on the right track.

In a sandbox, exploration itself can't advance the plot -- so instead, use a combination of the passage of time (that's the pressure) and player activity: meeting people, solving puzzles, making decisions, overcoming challenges. Change up the pace from time to time.

Sometimes James Bond is exploring at his own pace (he's master of the situation) and at other times he's desperately running away from bad guys (they're masters of the situation). Then he gets away from them or shoots them and he becomes master of the situation again. Of course, not every game has to use a lot of pressure. You can let the player have a very relaxed experience if you want to.

One question some designers ask is, "What if the player just goes wandering around and never seems to get on with the story?" The answer is, it depends on what kind of experience you want him to have. It might be okay to just let him wander around. I'd love to explore the countryside in the Far Cry games without getting shot at all the time.

Far Cry 2

On the other hand, if you want to push the player through the story, then you have to ask why he's just wandering around. If he's wandering around because he's lost or confused, that's your fault.

The designer Chris Bateman wrote a chapter called "Keeping the Player on Track" in the book that he edited, Game Writing: Narrative Skills for Videogames. In the chapter he talks about funneling: various tricks for helping the player find the "spine" of the game. In an open world you can't use the landscape to forcibly funnel the player back to the plot, but you can leave various signs and clues around. Get the book for more information.

On the other hand, if the player is just fooling around and you want him to get on with it, that's when you have to increase the pressure with player-independent -- and location-independent -- plot events.

Tabletop RPG game masters are very familiar with this. If the players won't come to the plot, bring the plot to them. If they've been hired to take down a crime boss and instead they're just sitting around the tavern gambling, the crime boss might get wind of their plans and send a gang of thugs to the tavern to squelch the expedition before it gets started. In the ensuing fight the tavern just happens to catch fire. Even if the party survives, it won't be doing any more drinking in there.

Another question people sometimes ask is, "In an open world, how do you prevent the player from seeing something early that he's not supposed to see until later?" The question is rooted in the assumption that everything that the story needs will be physically present in a static game world from the beginning -- as it usually is in adventure games and Western RPGs, where the story is mapped to locations.

But we're not mapping the story to locations, we're mapping it to time and player activity. The answer is simple: don't put an object in the world until it needs to be there. In the Grand Theft Auto games you can't destroy a car in Mission 1 that will be needed in Mission 3, because the car simply isn't in the game world at all in Mission 1.

You obviously don't want cars suddenly popping into existence in front of the player's eyes, but you can bring a car out of a (formerly) locked garage. The player can't be in more than once place at once, so you can do all kinds of things behind his back.

So what kinds of stories can we put into big open worlds that the player can explore in any order, and in which travel is an integral part of the experience? Well, here's a handful:

Find the buried treasure. This is a low urgency game good for kids because the treasure's not going anywhere. The player runs all over town assembling clues. The clues don't necessarily have to be found in a specific order; they might be scattered pieces of a treasure map.

Find the buried treasure before somebody else does. Same story, but there's some pressure on the player. You can make this a simple race, or you can raise and lower the tension by having the enemy team try to sabotage the player.

If they succeed, then the player is under more pressure; if the player sabotages them instead, some of the pressure is off. (Be sure this doesn't create too much positive feedback, though.)

Find the time bomb. Obviously the most pressure of all, and tricky to pull off in open worlds, but not impossible. One way to give the player a little control over the pace is to endow the avatar with a limited amount of super-speed, like Superman or the Flash or Neo from The Matrix. The player can use the power at his discretion to buy himself a little more time.

Find people rather than objects, and of course this is made more complicated by the fact that people move around. Players can be bounty hunters looking for fugitives, private detectives looking for missing family members, counter-intelligence officers looking for spies, and so on.

Police procedural. This goes all the way back to the old Sierra On-Line Police Quest games. To build a watertight case, the police spend a great deal of time traveling around to visit witnesses, question the known associates of suspects, and look for physical evidence. Some evidence can be found in any order, while other evidence appears only after following a chain of clues.

You might need to keep your town small, since although travel is an intrinsic part of the job it's not terribly interesting. You could spice it up a bit by letting your cops deal with street crime or spot witnesses or suspects walking around. To keep it low-pressure, have players search for evidence with a suspect already in custody. To add pressure, let the suspect escape.

Infiltrate a large open area from any direction. Too many shooters put the player on a rail. It's cheaper that way, but it's less interesting for the player. In the current Afghan war, the NATO allies have air supremacy and helicopters, so they can put troops down anywhere outside a combat zone and let them walk into it from any direction they prefer. Mission planning involves examining aerial photographs and choosing an approach that looks good.

Escape through hostile territory from somewhere in the middle to the edge. In my game design workshops I often challenge one team to make a game about the Underground Railroad, the network of abolitionists that helped escaping slaves to freedom before the American civil war. Some of the real Underground Railroad routes were hundreds of miles long, and thus not convenient for a video game, but you get the idea.

Smuggling is about not just infiltrating or escaping, but doing both -- and often with an awkward cargo. During the long, long wars between Britain and France in the 18th and 19th centuries, trade between the two nations was technically cut off -- yet there was still plenty of brandy in Britain. The coastlines of both nations have numerous inlets that a warship can't get into, and the smugglers knew them all.

Like police, reporters do a lot of traveling, usually under deadline pressure, to gather information. So do spies.

Hunters naturally move around searching for game. Hunting doesn't ordinarily generate much of a story unless something unusual happens, but occasionally it does. As a kid I read two different stories in which wealthy (and therefore evil, apparently) big game hunters indulged a secret passion for hunting human beings -- specifically, the guides they had hired.

Root out the criminal gang. Professor Moriarty's tentacles are everywhere. No matter where Sherlock Holmes goes in London, he encounters evidence of Moriarty's dastardly deeds. But where are Moriarty's senior lieutenants, where is his headquarters, and where is the man himself? 

Resistance. Another game idea I give out in the workshops requires the team to design a game about a resistance movement in an occupied country. They're not allowed to make it a shooter; the game has to be about sabotage and making the occupier's lives miserable without getting caught. This, too, can be spread over a wide space, with soldiers of the occupying army constantly searching for the resistance fighters and keeping the player under pressure. 

There are plenty of other kinds of jobs or hobbies that routinely involve travel: fire fighters, electricity linemen, tornado hunters, restaurant critics... not all of these are necessarily suitable for video games, but it only takes a little imagination. I'm sure you can think of more.

To make an experience story-like, you have to avoid too many repetitive or random (unrelated) events. (See my column Dramatic Novelty in Games and Stories for more about that.) If you read a thriller set in World War II, it doesn't consist of shooting an endless parade of identical Nazis; every situation is unique. This means that your sandbox has to be full of all different kinds of things, not just a lot of the same thing.

This is probably the strongest argument against sandbox storytelling: it's expensive and a lot of work. But unlike rail games, if you construct the world carefully enough, the game will be highly replayable. Different paths through the world will offer different experiences. Nor do they need to have the same objective or ending.

For several hundred years the people of Rome gave their allegiance to one of four factions that supported chariot racing. The drivers wore colored clothing so people could tell them apart, and the factions were named the Reds, Whites, Blues, and Greens.

Suppose the player has just moved into Rome. He can join any group of supporters, just as we can support any sports team today -- but with a difference: the factions often rioted, and there were bloody fights in the streets. What this means in practice is that an NPC who belongs to the faction that the player chooses is an ally, but if the player replays the game and chooses a different faction, the same person is an enemy. No need to write two stories or design the character twice; drama naturally emerges from the situation itself.

In short I think sandbox storytelling both possible and fun. You'll need to fill the sandbox pretty full so as to offer plenty of dramatic opportunities (many sandboxes feel rather empty and sterile), and you'll have to decide how much pressure you want to put on the player and how you'll apply it. This may include using some time-dependent, player-independent plot events to keep things moving forward. The environment itself is also critical -- it has to be a place that the player really enjoys being, because he's going to spend a lot of time there.

There was a famous film noir called The Naked City that was later adapted into a TV show. At the end of the film, and every episode of the TV show, the narrator said, "There are 8 million stories in the Naked City. This has been one of them." Try building your own Naked City, and see how many stories you can get in.