Wednesday, March 23, 2011

GDC 2011 Transcript: Motivating Players in Open World Games

Since I don't use speaker notes, my slides from GDC 2011 are basically useless on their own.  Because of this, I wanted to post this transcript of my talk online.  I typed the majority of this out on the plane to San Francisco, mostly as a way to practice the talk quietly, so it follows what I had planned to say pretty closely.  In fact, I realized through the week that I had forgotten to make several minor points which are captured here.

Fair warning - this is long.  Turns out I speak about 7,000 words per hour.  Transcript after the jump.


During the Level Design in a Day tutorial at GDC 2010, we took a lot of questions from the audience. One question directed at myself was: “How do you go about making levels for open world games?” Should be an easy one, right? The vast preponderance of my experience at this point in my career has been doing just that, after all.

I was stumped. I gave some answer that I felt was kind of a lame cop-out, and it bothered me for the rest of the conference. By the time I left San Francisco last year, I knew that this would be where to begin if we were invited back in 2011.

We did get invited back, and I found that the question had not gotten any easier to grab hold of. It was just too broad. There was an implication, too, that conventional level design wisdom didn’t apply, and this wasn’t true – although many tricks that served me well in more traditional games had to be tweaked for Open World efficacy. In fact, I realized that building for an Open World required some small, but fundamental shifts in how I viewed my role as a creator. Much of this was something that had happened gradually, subconsciously over years at Bethesda. So I started thinking about this notion of building for player experience in open games specifically.



The first thing I wanted to do was consider what we mean when we talk about an “Open” game. I didn’t want to attempt some one-size-fits-all definition, though. I don’t think of “Open World” as a genre, for one -- just a descriptor you can apply to many types of game. Shelves are lined with dozens of titles we might apply this label to – and certainly there’s been an explosion of Open Worlds since GTA III got the world’s attention with big sales and genuinely captivating gameplay. The notion of an Open World game isn’t new, though. Look further back and there are many classic titles you can apply the moniker to. Freelancer is a personal favorite, and many RPGs like Arcanum had expansive, non-linear overworld maps before GTA III brought the concept to the mainstream. Even the early consoles had them – Toe Jam & Earl, Final Fantasy, the original Zelda. Speaking of which, it’s not an entirely Western or American idea, which is sometimes the misconception. Games from Asia – Ico, Ocarina of Time, Symphony of the Night – as well as other parts of the world – Risen, Mount & Blade, Minecraft – have explored this territory.


Further, another term that often gets thrown around when we talk about Open World games is “Sandbox.” Whereas “Open World” is just a statement about the physical space, “Sandbox” is often used in a way that could be interpreted as “Open Mechanics” or “Open Simulation Boundaries.” I don’t think many people would argue that Deus Ex is an Open World game, for example, but it has this same feeling about it; you can choose to approach the setups in the game in many different ways.

This, for me, boiled down to one common aspect -- a deliberate choice to focus on Player Agency.

So why is player agency attracting so much attention? Why bother talking about it? What are the roots of our fascination with it? For me, the answer would end up being personal. I have a memory from my early childhood. The details are vague – I don’t know if it was an Atari, a computer or an arcade cabinet – but the emotion is clear. I remember feeling awe when I first saw something on screen manipulated directly by an external controller. This flew in the face of everything my young mind knew about how a TV worked and how you interacted with it. Suddenly the TV offered playtime, not passive consumption. This same draw is what has spurred a generation – and, increasingly, our global culture – towards interactivity. Even in today’s world, where we’re surrounded by interfaces and interactivity constantly, I still find myself occasionally inspired when I see something on-screen happen when I press a button, if just for a passing moment.

I won’t be the first person to point out the obvious fact that Interactivity is the primary, and perhaps only, attribute of games that other mass media hasn’t got. It bears repeating, though. We may lag behind film, for example, in story-telling, but I believe we can catch up one day, and that’s a noble pursuit we need to focus on. But Interactivity – the fundamental core of our form – they will never have. This is why I believe in games the way I do. If I didn’t, I wouldn’t be in this industry – and I believe I have gravitated towards Open games for the same reason. Games that choose to focus on interactivity are the path to unlocking the full potential of our medium.  


It presents problems for us as creators, though. We have a new collaborator whom we’re inviting into the creative process. The player may not directly inform creation, but games, like any storytelling medium, come alive in their audience. Games, especially, have to trust the audience with a lot of the quality of that experience. You can put a book or a film down, but you can’t usually influence the directing or script. And with Open games we’re choosing to give the player even more autonomy than usual. The control we give the player is inversely proportional to the amount of authorial control we are able to exert. This is just a reality, however. We simply need to acknowledge this partner and accept the relationship.

Let’s pause for a second to consider a specific type of player: the type that gravitates towards Open games. These are players who love to prod and explore. They want to experiment with cause and effect, to see how the game reacts when they do this or that. These are players who pick up on design cues – go left here – and do the opposite. Other games have trained this behavior in part. Lots of games have trigger events that advance a story, after which you can no longer go back and explore the space you’re currently in. These explorer types may be avoiding that, or just be trying to be sure they’ve fully explored an area before moving on.

This type of player can be trouble, and designers tend to build a lot of fences around their content. Invisible walls, contingencies in scripts, locked doors – usually with a well-intentioned goal of protecting the player from ruining her own experience. The problem is that players will always find a way around these band-aids. In my experience, if you’re simply reacting to issues like this as they come up in the form of bugs, then your content is going to become bloated and difficult to edit – and you’ll inevitably miss something, especially when your player pool goes from 50 QA testers to however many thousands or millions of people play the game. Simply accepting this will often inform your process to use simple, open handling – and increase the chance that when the player breaks your handling, she doesn’t also break the game.

The most important thing to remember is that this player isn’t a jerk. Every designer will have this reaction from time to time. “The player who does X is trying to break the game, I’m not handling this edge case!” Well, maybe. Probably not, though. These players are our people. You probably play games the same way they do. We aren’t building delicate model airplanes; we’re making Tonka trucks. Our games are meant to be banged on, played with, and not break too easily if we can help it.

So – back to this player/designer dynamic. Like any relationship, things are going to go more smoothly if there are some ground rules.
  • Player-Designer Ground Rules

The first rule to remember is that the player is ultimately the one in control. This may seem obvious out of context. After all, we aren’t on the couch with the player, and all of our responses to player activity are preconditioned in some way. Yet we still have this illusion as creators that we have ultimate deterministic control over gameplay, when this just isn’t the case.

Consider Jason Rohrer’s “Sleep is Death,” a turn-based storytelling game for two players. One player is the storyteller, the other simply the player. Each turn, the player is able to move around a simple 2D area and interact or speak via simple text boxes. The Storyteller then has a window of time in which to react to the player. The storyteller presumably has an existing story in mind, and has various prepared bits of writing, background sets and other assets. The dynamic inevitably ends up with the player really driving the experience, while the storyteller scrambles to keep up – frantically, if the player isn’t being cooperative. In fact, when a SiD storyteller tries too overtly to funnel the player, the experience becomes less enjoyable – the spell is broken. It’s a lot like D&D; the Dungeon Master does a lot of prep work, but at game time the players are really the ones in the driver’s seat. At the end of the day, a common complaint for an unsatisfactory game is a heavy-handed DM or SiD storyteller trying to constrain players to follow a preordained story path.

Which brings us to our second rule: Stay backstage. Strive to be unseen. We like to keep a 4th wall between ourselves in the player. If we think of the player as an equal partner when we are creating, that’s great. The player, however, shouldn’t have the sensation of our constant presence. Your best work will go unseen – the game will just feel like it’s doing what it should, reacting to the player naturally. Remember: Tonka trucks. Try to build for open play from the start.

Third is fulfilling expectations. This is one that absolutely applies to all games, not just Open ones. This comes down to consistency and trust. The player should have a reasonable expectation of what to expect from various scenarios. For instance: these quests give armor rewards; this tactic is effective against that enemy; breaking a ? box gives me some kind of power-up, or if I follow this path I’ll find something fun. It also applies to simulation rules of the world. If the player can kick down a door in one level, it’s going to feel bad that a dozen other, similar doors in another level cannot be. This game-wide consistency builds a foundation of trust, and when reasonable player expectations are not met, it can often be disappointing and damage that valuable sense of trust. I should offer the caveat that playing with expectations can sometimes be refreshing, but it’s something to be careful with.

Fourth, and most important, is to accept that the story in the player’s head is always more important than the story you thought you were telling. Check your ego at the door. There’s a lot of talk about player-driven experience and emergent narrative. Whether or not you personally buy into it, this is important to many people who will play an Open game. This might come off as an anti-story comment, but it isn’t. The storytelling potential of games is exciting and far beyond what we’re doing right now. What I am saying, however, is that when we are trying to tell story, the player must feel like it’s occurring naturally around, or in response, to her. We shouldn’t just be telling our story at the player.
  • Directing the Player to the Fun Stuff


With these ground rules in mind, let’s confront one of the first, fundamental problems they present a level designer. My broadest definition of a level designer’s job is to ensure that the player is entertained. The definition of “entertainment” might shift to reflect the game, but that’s the basic responsibility. However, if we just got finished saying that the player shouldn’t feel our presence, how do we make sure she is entertained? Well, imagine giving somebody $50 and dropping them off in a rough part of town, versus dropping that person off at the State Fair. Now, the person downtown might have fun – more fun than you’d ever have at the fair, even – but they might also wander aimlessly or get stabbed. The person at the fair will almost certainly have a good time, because it’s an environment set up for fun.

Fun in our games is usually concentrated as some kind of POI – point of interest. These are the places in the world where we’ve put effort into setting up interest, and the primary tools in the box for any level designer or world builder to draw players naturally to them are simple environmental techniques.

First, and most obvious, are distant visual landmarks. The classic example here is Cinderella’s castle, now a standard feature copied by themeparks everywhere. It’s one of the first things anybody sees arriving at the park and guests naturally head towards it. The interesting part about Cinderella’s castle in particular is that it’s not a ride or show. Primarily it’s a landmark to help you orient yourself in the park. It’s also a planning hub, because from the area in front of it you can see many other landmarks you might have missed before, sprinkled throughout the park.


These other landmarks are basically doing the same thing as the castle, but on a local scale. Consider the screenshot from Fallout 3. These satellite towers are the main distant landmarks in this part of the Capital Wasteland, but check out the little landmarks nearby. There's a small factory, and also a water tower will become visible as the player nears that initial landmark. We’ll talk about this kind of distraction and its utility later.

We also use these kinds of landmarks in the Point Lookout DLC to set up several landmarks to draw the player into the new region right away, as you can see in the video below.  The lighthouse and the ferris wheel are just cool weenies to draw interest.



The smoke rising from the mansion and the enemy silhouetted at the start of the shot both provide some movement to draw the eye, too. You can get a lot of use out of motion in the environment. Movement always creates more visual interest in your space, but it also tends to have a direction. Streams flow this way, a flag in the breeze flaps that way. These directional cues can guide the player along, some more powerfully than others. When we build POIs for Skyrim we can keep these cues in mind. For example, we’ll tend to put something interesting at the pond where a creek ends, because we consider it likely that players will follow the stream to see where it goes.

Another tool at our disposal is prior knowledge of the world. Put another way – use common sense. Games have a lot of conventions – exploding barrels, the need to break open crates – that don’t have a strong counterpart in the real world. These are fine, but remember that even hardcore gamers were unaware of them before they started playing a lot of games. Whenever we can use knowledge which is common to the human experience, we increase the efficacy of our efforts to guide the player along. Roads and signposts are a good example of this, as are signs of life, like the smoke of a distant campfire or the sounds of civilization.

It's also important that we remember the power of sound. When we play games, we typically only have two of our senses at our disposal. Vision is already doing the heavy lifting. Probably the best use of sound to draw players in Fallout 3, as it happens, was an accident. As the player explores the world in that game, we’re constantly streaming in new area ahead of them. This means AI that was running in the background will start processing in realtime, and random encounters will spawn. This also meant that combat breaking out on the edge of the loaded area was quite common. This was the same as it was in Oblivion, except that Fallout combat is loud. What we observed was that players would be walking along in the wasteland and hear muffled, distant gunshots and explosions, and almost always turn on a dime to investigate. Lucky for us, this usually brought the player to something interesting we had set up. We’ve taken that insight with us into Skyrim, a world more visually dense than the Capital Wasteland, where players can easily walk right past a POI we set up without seeing it. Now we’re cognizant that the crackle of a campfire, the crash of a waterfall, or the ringing of an anvil might draw you to what you can’t yet see.

None of this is revolutionary stuff, though. Most level designers and world builders are doing this, even if at an intuitive level without realizing it. What’s really going on, though? Why does it work when it works?

Well, we’re trying to elicit a specific emotional response. We want to motivate the player to go somewhere. We’re trying to create goals in the player’s mind.
  • Goals and Player Motivation
And this is interesting, because I think it’s the building block of sustained interest in any given game. I think that the player should always have a goal. This isn’t something I’ve always believed, either. I play a lot of games without focusing on the goals presented by it. Far Cry 2, for example. Sometimes I just want to be in that game. I’m not doing missions when I play, though. I’m roaming around the hills, swerving to avoid zebras with my jeep, or torching a hostile campsite.


But I also drift away from most games that I play. I don’t fail to finish games because I throw them down in a rage, or even get sick of them. I just realize one day that I haven’t touched the game – maybe a game I really enjoy – for a week or two, and I’m kind of done with it.

What I realized in hindsight was that, in every case, I had ceased to feel compelled to do anything in the game. I had no goals, whether my own or those of the designer.

Goals can take many forms, then. The most obvious are the stated goals. When you start Fallout 3, for instance, you have an explicit quest goal to find your Dad. Now we know, because we’re comfortable with the fact that our story goals may not align with the goals or interests of the player, that the player probably isn’t going to off and find Dad right away. These stated goals are useful, though, because they provide some direction to the player. Especially for players who have become bored with the game and need something to do, or players that have put the game down for some time and don't immediately recall what they were doing with it.

There are also player-determined goals. Really, any goal can be player-determined. That is, if you really want to find dad, then fine. That stated goal is also the goal of the player – great. There are also goals which may not be stated, but exist because a designer constructed them. For instance, I don’t think anybody ever tells you in Crackdown to collect all the agility orbs. But they’re there, in interesting places to go, and many of us will do nothing but hunt them down for big chunks of time. Then there are goals that are entirely unscripted. Some of them we may still kind of create – we know that stealth players will gravitate towards certain setups for an easy kill, for instance – but some are completely out of left field. Somebody playing an MMO may roleplay a blue knight and do unusual things to acquire the coolest blue armor pieces in the game. We know full well that somebody will play Skyrim and try to kill every elf they see, too. The important thing is the player choice to focus on a goal the player cares about.


There are also emergent goals to consider. The term “emergent” gets thrown around a bunch in different contexts, but what I mean here are goals that occur out of the simulation rules of the game, in pursuit of some higher goal. The best example I can give here is Minecraft. Minecraft is an interesting game to bring up, too, as it’s a game that currently has no stated goals or authored narrative whatsoever. Let’s say a few Minecraft players tell us their goal in the game. One wants to build a portal to the nether. Another just wants to explore a cool environmental feature. Maybe you want to build a cool rollercoaster like you saw on YouTube. Take the rollercoaster example. If you’ve played any Minecraft, you know that this is going to require a ton of iron. You can’t get that kind of iron without a mining operation, though. You need tools to mine, so you start clearing a forest, and you need to build a little house to keep you safe from monsters at night, too. Then you find this underwater river and end up finding a monster spawner and…. You’ve been playing every night for a week, and haven’t laid the first piece of track for your rollercoaster. You don’t throw the game down in frustration at this point, though. You’ve been highly entertained by all these little emergent goals that bubbled up, and may not even care about the rollercoaster anymore. This has everything to do with why Minecraft is so good. The simulation rules are just very well done, and create a nice pace of experience.

Somewhat similar are Goals of Opportunity. These are goals that pop up and tempt the player for a moment. Say you’re playing WoW, and you notice something on your mini-map. There’s a mineral deposit nearby – something you could really use, not far away. You check it out and sure enough, it’s in the middle of a camp of monsters just a bit too high level for you. The player makes a value judgment on the fly. Do you try to carefully pull each mob, expending resources, old school style? Maybe you call on a guild member or try to stealth in and hope you get the ore before they aggro you? Perhaps you’re feeling sneaky and wait for another player to come by and draw aggro, allowing you to ninja-loot the ore. Or, maybe, you keep riding on by. These are great little decision events that keep players interested.

Finally, and a little bit less relevant, but still interesting, are out-of-game goals. Trophies and achievements have given us useful way to communicate very “gamey” goals to the player. Pretend that we think there’s something really fun about getting 100 headshots, or killing ten enemies with a single explosion. That’s difficult for us to do in the narrative context of the game without breaking the fourth wall. This external layer, however, is a chance to speak, designer directly to player, and offer these goals without breaking immersion.

Given two players of the same game who happen to have the same goals, however, it’s unlikely those goals are prioritized in the same way. Different goals appeal to different players and different circumstances. We can begin to understand this judgment by breaking down that goes into it a bit.
  • Goals & Priority
The component of prioritizing goals that any designer will know about is the risk:reward ratio. Recalling that we want to fulfill the player’s expectations, we are hopefully able to telegraph the difficulty of a given scenario to the player, so she can surmise the reward. A player might see that a platforming puzzle is difficult, or notice evidence of a challenging enemy outside the mouth of a cave. Maybe your game has a direct way of conveying both risk and reward, like WoW’s quest journal. Players will gravitate towards rewards they want and risks they feel they can handle. This is a subjective value judgment.

Somewhat less obvious is the notion of commitment – time commitment, specifically. I would hypothesize – I have no data to support this one way or another – that players tend to gravitate towards a goal they think matches the time they have to play. For instance, if Joe Gamer sits down on a busy Wednesday night, he may only have 15-30 min to play; he’ll probably choose something he thinks he can finish in the time frame. Likewise, if he’s playing Fallout 3 on the weekend and has a few hours to devote, he may decide this is the day he finally finds Dad.

The real monkey wrench is inherent interest. There’s truly no accounting for taste. We probably have a good idea that certain things attract certain player archetypes, of course. The player who is into dialogue and bartering will gravitate towards towns, while Conan is going to grab a broadsword and go dungeon-diving. Sometimes there are player goals we can’t know a thing about, however.

When we announced Skyrim a few months back, I was lurking on the forums and saw a speculation thread about spiders. The poster, it turned out, was curious if we’d have spiders as an enemy type in some dungeons, and if we’d include an option to disable them. I thought this was silly at first, but then I skimmed the thread and saw that other people chimed in with the same sentiment. Then it dawned on me – I know one of these people.  


Last year I got back into WoW for a bit, and convinced my friend Casey to try it. Casey plays some games, but she isn’t the biggest gamer in the world by a long shot. I remembered that when I was helping her level her character, she would always avoid spiders. I know Casey, and she doesn’t have crippling arachnophobia. She just thinks they’re creepy and would rather not deal with them. As a lifelong gamer, this was enormously frustrating to me – if spiders give me the best XP, or if I have a quest for them, or even if they have a 1% chance to drop something I want, then I’m killin’ spiders. I don’t care if they’re unicorns and teddy bears, I’m going to get genocidal on that monster. What I realized, though, was that I was playing WoW like I was balancing spreadsheet. I’m pretty conditioned in a game like that to do whatever is the most efficient thing possible. Casey wasn’t enjoying WoW any less for her preference. That game has a broad enough scope that we could easily go around the spiders and do other stuff – and sometimes that actually made the experience more interesting.

Of course, it isn’t as though we sit around at Bethesda and chart out this kind of stuff for every quest and POI in the game. What’s the point of discussing it, then? If player choice is so chaotic that we can’t accurately predict behavior, why bother? Well, it isn’t about building a perfect metric. It’s more about awareness of this fact, and being able to cope with it as well as possible.
  • Deliberate Distraction

In preparation for this talk, I spoke with Rob Davis, who worked on The Saboteur. He said that the golden rule for him in Open world games was to provide the player with as many interesting things to do whilst doing from point A to point B as possible. That’s spot-on.

It’s something I call deliberate distraction. Let’s take that A to B example for a moment. It’s a Common setup in any RPG and most open world games. Let’s imagine that as a graph for a moment though, where the vertical axis is how entertained the player is, and the horizontal is the time spent playing the game. We’ll go ahead and assume that what happens at both A and B are pretty interesting things, but that doesn’t mean the player interest level is static between them. In fact it almost certainly wanes leaving A and falls into this valley of boredom.

So let’s revisit that scenario. We still have A and B, but we sprinkle some points of interest between them – X, Y and Z. We actually do this with quest paths at Bethesda. We’ll draw lines along quest routes and use that to inform where we place POIs. Some designers will be uncomfortable with this. You might worry that the pacing of your story will fall apart. Remember, though, that the player story comes first. If your story has a player fully compelled, chances are they’ll bee-line. Be comfortable with the fact that they may not, though. Don’t shun that player.


Let’s just say the player checks out X and Y, but not Z. She didn’t notice Z, was in range of B and wanted to bee-line, or maybe just missed it. No big deal, it happens all the time. Perhaps there were spiders involved. When we take that journey and plot it along this imaginary graph, notice too that X and Y are a little lower on the vertical axis. Let’s just say they were less interesting than A and B. That’s fine, too. Not everything will be a big set piece. When you chart the player interest, you see that we have less sustained time spent in the valley of boredom. It doesn’t matter that X and Y weren’t mind-blowingly awesome; they still contributed some entertainment and improved the overall pace of the player story.         

Deliberate distraction is just one way we cope with the next problem we want to confront – empty space. Proper use of physical space in any game, but especially an open one, is critically important, and a big part of the level designer’s responsibility.
  • Dealing With Empty Space

Let’s pretend we have a job – remake God of War. Not a sequel, not a reboot – we’re remaking the first game, but this time it’s going to be an Open world game. Well, that game has a lot of these dramatic backgrounds, like polygonal matte paintings. So that’s what we’ll focus on. Bang that out, make all the space playable. You see it, you can go to it. Cool, right?

Well, no. If we add nothing to those areas, except that now the player can run off and explore them, we haven’t added anything of value to the game. In fact, we’ve just made the game worse. Because if the player wants to check something out – maybe some temple reminds him of a vacation spot from childhood – and there’s nothing there. What was the point? Poseidon is still wrecking things back there. That’s still where the fun is.

This is another point where some developers will become uncomfortable. There’s an attitude you might encounter in yourself or others when confronted with this. “That’s just for looks. The player isn’t really meant to get there.” That isn’t good enough. You can’t just abandon those areas. You made it playable, and remember that a good chunk of players will go explore that kind of stuff, hoping to find something neat. What they’re doing is totally valid.

Take another example. The Rope Bridge is a classic adventure story setup. There’s always the tension – will they make it or will they fall? In the Temple of Doom – spoiler alert – Indiana Jones and friends make it across just fine. And if we were to make a Temple of Doom game with the goal of recreating the story beats from the movie, that would be all we have to handle. But we’re making an Open world game in that setting. Now we have to handle something – namely, the fall into that river full of American Alligators posing as crocodiles.

For the linear game we didn’t really have to worry about them. Indy either couldn’t fall, or we would play a cinematic of the gators thrashing about and reload the scenario. That won’t do in an Open world game, though. So what to do? First we have to actually build the area to make it playable. No smoke & mirror tricks, we’ll have to physically connect it to the world. But what if the player falls?

Your instinct may be to place a killbox in the canyon. Which you can do, but that sucks. That’s something we really try to stay away from. People don’t come to open world games for that. “Okay,” you say, “it’s a really long fall. The player’s probably going to die anyway.” That might be true, but remember that the player will always find a way around your little blockers and barriers. Somebody will slide down the edge safely, or have enough health buffs to survive it. So that’s not really valid – and it’s functionally the same as putting a killbox in there anyway. You may try another trick: “Those crocs are beasts. They’re going to rip the player to shreds, and the player’s going to get it. You shouldn’t be here. Go back and cross the bridge,” and at this point you’ve laid down the gauntlet. You’ve just issued a challenge to the player, and the player is going to have the skill, items, or old-fashioned patience to get through the fight and wipe out every croc down there, one way or another. And there’s no worse kick in the teeth than the fact that you’ve got nothing down there for the player but a middle finger.

What do we do with that river, then? Well, it’s the reality of building such big games that we can’t have something incredibly interesting in every nook and cranny of the world. It’s fair to point out that we spent a lot of time on whatever was across that bridge. Which makes sense -- it’s something we expect the majority of players to see or do.
  • Consolation Loot

One thing we can use here is what we at Bethesda call “Consolation Loot.” Basically, we’ll put something – anything – in these kinds of areas. Consolation loot can be many things. We might use a freeform quest hook, some optional story bits like a journal, or the most straightforward solution – some minor loot. Some designers may suggest that we’re rewarding failure, but remember that many players will deliberately investigate these nooks and crannies, and for those people you’re rewarding exploration. Having something in these corners will make the game feel much more handcrafted and polished. We don’t want to put anything critical to progressing through the game in a dead end, but even if the loot we put in that dead end is inconsequential, it still takes a moment of zero interest – or frustration if the player went through duress to get here – and turns it into a positive moment, however minor.


It’s not just a Bethesda thing, either. If you’ve played any Final Fantasy to completion in the past 15 years, you may have ended the game with Hi-Potions or Phoenix Downs maxed out. Even though some of those games are fairly linear, they’re filled with dozens of little dead-ends where you’ll find these loot items. They may be inconsequential in the grand scheme of things, but that doesn’t matter. What matters is the minor joy of finding them.

Let’s take a step back and look at this problem of empty space on the global scale. Read reviews or forums of any big, Open world game, and a common theme will come up again and again: the game feels vast for the sake of being vast. “This is a walking simulator. There’s no reason for this game to have such a big world if it’s empty.” I believe there’s only one way to overcome this hurdle, and that’s by paying attention to your POI density.
  • POI Density

POI density, simply put, is the notion that if we choose a point in the world at random, we have an expectation of the number and variety of Points of Interest in the general vicinity of the player. I can’t give you a perfect formula for POI density, but I can give some basic ideas on how to determine and use it.

First, figure out what POIs your game should have. They can be tiny campsites, outdoor ruins, dungeons to explore, whole cities and towns, or non-physical things like randomized, roving encounters.

Next, decide how long each type of POI should take to create. One good level designer may knock out a half dozen minor landmarks in a day, whereas a fully-fledged town could take multiple people the span of a project to implement.

From this point, you should be able to develop an idea of your desired POI density. This will be influenced by many factors, such as draw distance, travel options, visual blockers in the world, and tone.

Imagine a hypothetical area of your game. Let’s say we’ve decided that a square quarter mile is a good feel for the immediate area surrounding the player in your game. Now begin to theoretically populate that area with the correct number and variety of POIs to achieve your desired POI density. We can think of this as a sort of “Vertical Chunk” of the game. By reconciling those chosen POIs and the time it takes to build them (and the world they occupy) we can theorize how long that Vertical Chunk would take to build.

This is the part that none of us are doing when we approach our Open Worlds. Most games have a relatively known timeframe in which to be built. With our vertical chunk estimate, we can extrapolate to figure out how much world we can build in the time we have for development. Yet most Open World games begin with a fixed world size, and the teams are simply doing their best to fill them up with fun. It’s a trap we fell into with Fallout 3.

One of the first and most valuable pieces of documentation we had for Fallout 3 was a map of the Capital Wasteland. This was an invaluable planning tool. This got everyone on the same page about the general size of the map, what locations needed to be built, and gave us a rough idea of quest flow; it provided a foundation for almost every decision made at the high level.

Fast forward nearly two years. We’re in alpha, and for the first time, we’re meeting Fallout 3. Up until this point, the game has been a collection of assets, tasks, code and lots of teamwork. The game is beginning to feel complete as everything comes together. We realized, however, that wandering the wasteland didn’t feel right. Something was off about that experience. What we realized was that the world felt pushed together. We simply had too much stuff, too close together.

This is because we were using the same POI density that had served us well for Oblivion. There’s no secret playbook at Bethesda that tells us what to do. We simply learn from what we’ve done before. POI density wasn’t something we really had a name for back then.

The trouble was that Oblivion and Fallout 3 are very different games. The sightlines in Fallout 3 are much larger, and you don’t have the major visual blockers of Oblivion’s forests and hilly terrain. Aside from all that, it’s a game where you’re supposed to occasionally feel lonely, and we weren’t achieving the tone we wanted.

So we made a decision. We could have pared back, but we decided to add a significant amount of new area to the North and West ends of the map. Remember that we were in alpha, and a lot of this work was supposed to be done. If you’re a producer, you should be squirming in your seat right now. This took our entire environment art and level design teams offline for the better part of two months. It was a painful decision, but we felt it was the best option to make the game what we wanted it to be.

The core lesson here is that you need to respond to your game as its being developed. Our process is deeply iterative, but at the end of the cycle we put a stamp on it and ship. The player will never see any iteration but the state in which the game is at ship. We might update a game through patches or DLC, but nothing on the scale of what I’ve just described. We think of our world as the main character of the game, and like the player it can be a fickle, evolving creature that we don’t always understand.

And that’s another important relationship the player has. Players who talk about an Open World game they care about tend to have strong, almost emotional ties to the location where the game takes place. There’s a real emotional resonance there, like I mentioned with myself and Far Cry 2 earlier. You just like being there. Being there has become its own, far-reaching goal. Powerful stuff, really

And we're a part of that as designers. By seeding the world with well-made locations, and thinking seriously about their distribution, the player enjoys a consistent, entertaining pace of activity and interest in the world. The world should feel at once believable, but more exciting than the mundane world. This is a place where adventures big and small are likely to happen; where you can trust that you’ll be entertained at some level. You aren’t stepping into the streets of your home town – you’re going somewhere more. Someplace that has spoken to you in a deep, compelling way that, as the player – even as the designer – you may not always understand.

3 comments:

Tim Flemming said...

Thanks for putting this up for people like me who couldn't make it to GDC.

Was a really good read and covers a lot of the questions I've been asked in interviews recently (about how stuff I've done in single player design will translate to MMO design).

I think your point about learning to estimate your schedule based around POI density is one of those ideas that makes a ton of sense in hindsight.

As an aside, I actually really liked the high density of a lot of Fallout 3, it reminded me more of Morrowind vs Oblivion. However, I think you're right in that to meet the desired tone of the game, the more open areas were a good addition.

It seems like playing with and varying POI density within the game may even be more desirable than trying to maintain a constant density throughout.

Thanks again for the insight!

Apples said...

You mentioned on Fallout 3 that, "it’s a game where you’re supposed to occasionally feel lonely, and we weren’t achieving the tone we wanted."

I'd like to know why you felt making an occasionally lonely game important in terms of the entertainment a player would have.

For me unfortunately, it was during those lonely parts of walking around in the map that ended up boring me. Sure I got the tone of a nuclear apocalypse, but for me it was at the cost of entertainment. I ask because I wondered if this method made the entertainment value go up for more gamers than it detracted from.

For me, who felt bored by these long stretches of walking, I recall having much more entertainment value with Fallout 1 and 2's method of giving the sensation of vast stretches of land between points of interest, but keeping travel as quick to the point as possible. From there, most of my time was spent at various points of interests rather than the monotonous act of walking from one place to another.

Thank you for the article.

-Andy Kim

inspire said...

Joel Burgress! Could you help me with a few questions? I inspire to be a game designer and thinking of a college to go to for it. I want to know if the college I pick is the right one and any pointers you could give me. Thanks and was Thinking Full Sail University online for game design. Thanks