The Making of Dust: Architecture and the Art of Level Design
What makes a room, house, or city block livable? Does the answer have anything in common with what makes it playable? In crafting game worlds—“maps” or “levels,” as they are called in the business—a designer creates spaces that on the surface may seem as close to reality as technology allows. In fact, game designers often emphasize the exact opposite of what makes real-life spaces inhabitable. David Johnston is the creator of Dust (de_dust)—one of the most-played maps in history. Here, he examines what has always made his work so popular, and why successful level design is the antithesis of successful architecture.
In the spring of 2014, gamers from across the world arrived in Katowice, Poland, to witness one of the biggest video game tournaments of the year. Among them were sixteen teams of professional gamers who would be competing in a $250,000 tournament playing the online first-person shooter Counter-Strike: Global Offensive.
This wasn’t the first tournament of its kind, nor the biggest, but it was indicative of a growing trend. E-sports, once only a staple in countries like the United States and South Korea, had gone global. Spectators arrived from every corner of the globe, while even more of us watched the event live online, accompanied by full pre- and post-match analysis and commentary not entirely dissimilar from mainstream sports events. For the competitors, victory would require determination, fast reflexes, and smart tactical planning as their teams were pitted head-to-head in a series of matches in a handful of virtual 3D arenas to determine who would be flying back home with the loot.
One of the maps being played—known simply as Dust 2—I had designed nearly fifteen years before. It was the successor to Dust, a map I had made just a couple years before that on my parents’ home computer, instead of studying for high school exams. Dust was set in a sun-baked, ramshackle town of an anonymous Middle Eastern country, pitting a team of terrorists armed with an explosive charge against a team of counter-terrorists racing to stop them from planting it in one of two locations. In tournaments like the one in Katowice, two teams play fifteen three-minute rounds then swap sides, and the team that wins the most rounds is declared the winner. Dust had proved to be reasonably popular at the time, and I had been egged on to create what I thought would be a lousy sequel.
I was wrong. So terribly, irrefutably, embarrassingly wrong.
Dust and Dust 2 are now renowned as two of the most iconic maps of online and competitive gaming, arguably two of the most-played maps in gaming history, and most definitely the highlights of my entire level design career (before it had even really begun). Dust 2 in particular is a regular appearance at these competitive sporting events, with the original more popular among casual players of the game.
Unlike traditional sports, Counter-Strike doesn’t rely on symmetry to achieve balance. The terrorist and counter-terrorist teams have an array of different (albeit very similar) weapons, and are dropped in realistic 3D environments with entirely opposing goals. Level designers can’t create half a map, flip it over, join the two halves together and call it a day, because the different team objectives would make a perfectly symmetrical map unfair. While in competitive tournaments teams switch sides half-way through, in casual matches this isn’t the case. Designing a good, balanced Counter-Strike map that can stand a multitude of variables and the test of time is incredibly difficult to do.
Yet, somehow, I’d managed to do it not once, but twice. Both maps were not only in tune with the game, but instantly struck a chord with players. I should have been elated—but I wasn’t. I couldn’t be. The maps had proven successful, but I didn’t know why. Why did people enjoy them? And keep enjoying them? What exactly had I done?
Foundations
As a kid, I was fortunate enough to be brought up surrounded by ever-increasing mountains of Lego. I was fascinated by the facsimiles of buildings and vehicles I could create and spent weekends trying to replicate the things I knew and loved. I was hooked from the moment my dad showed me how offsetting each row of bricks in a wall would stop the wall from falling down. These were simple principles, and I loved exploring them. But it was all just preparation for what was coming next.
In the early 1990s, Wolfenstein 3D and Doom gave me my first taste of level design. Enthusiasts had released applications that let fans of Wolfenstein 3D create their own maps for the game using an interface not entirely dissimilar to the highly-regarded office productivity application, Minesweeper. By clicking and drawing on a rigid 2D grid of squares, we could populate a world with square walls, square doors, soldiers, secrets, and weapons, hit Save, then load our creation into the game to play it in glorious 3D. Too many enemies? Not enough weapons? Go back to the level editor, tap a few squares, and play it again. The fact that this 3D world was formed of precise 90-degree angles and bore little resemblance to the rich, flowing, non-angular, natural world as we know it was gleefully glossed over by our imagination.
Doom went many steps further, freeing us from the angular grid of Wolfenstein 3D and allowing us to create rooms of any shape and—a crucial new ingredient—any height. In Wolfenstein 3D, ironically, the game was little more than a flat 2D world drawn in a way to make it look like it had depth. In Doom there were stairwells, balconies, walkways, windows, deep, dark chasms, and towering monuments bathed in sunshine. It wasn’t quite full 3D yet (closer to 2.5D), but it was more than good enough to give a sense of place and being.
I rapidly became addicted to Doom—not for the visceral blood, guts, and violence it offered—but because the creators provided fans with the tools to create their own intricate, interactive 3D worlds for it. Lego had let me create small machines and houses brick by heel-piercing brick. Doom gave me the power to create vivid, interactive environments in less time than it would take to entice those stubborn pieces of moulded plastic apart again.
Level design became a bizarre form of escapism and virtual tourism rolled into one. Many of the first maps I created weren’t demon-infested space stations or war-torn military bases, but places I already knew from real-life—starting with the classrooms and hallways of my high school. I strived to get every detail exactly right, even with the distinct limitations of Doom’s technology. Doors had to be the right width, ceilings the right height, windows and railings exactly where they should be. I started to admire and appreciate the thought and execution of the design that had gone into the classrooms and offices I’d never really thought too much about before.
Some of the most fun I had was trying to fill in the gaps—replicating the parts of my school that I wasn’t familiar with, like the staff room where my poor teachers could snatch brief moments of calm between sessions of educational torment. As tempting as it was to fill this room with legions of blood-red Cacodemons and goat-horned Barons of Hell that would hiss and attack as soon as I peered inside (and, let me assure you, this was extremely tempting), it was somehow more satisfying to create a place that was in keeping with the rest of the world, down to the rhythm of window placement, structural pillars, and fire exits.
It’s hard to imagine that the mishmashed dated buildings and corridor mazes of my high school might have contributed in any way to a map that would be played by professional gamers many years later. Look at them side-by-side and Dust doesn’t look like my old high school—or indeed, any high school that I’ve seen—at all.
But if you look past the breeze-blocks, blackboards, cracks, crates, carpets, and sand-covered plains, then similarities emerge. Look harder and you discover that good level design draws heavily not just on good game design, but on the same principles used in civil design.
Little did I realize that in copying my school down to each pillar and doorway, I had been unintentionally learning the craft of the building’s architects—the logic they had used to construct a busy, working school became entwined with my growing technical knowledge of level design. Doom heralded a world where game settings were evolving from arbitrary, abstract mazes into rich, relatable environments. Meanwhile, level design was leaning more on established design principles than ever before, and we were learning how these principles could be exploited.
Spaces
At a low level, designers are focused primarily on accessibility and accepting simple physical truths. In the real world, we take for granted that floors will be flat, ceilings high, walls vertical, doors tall and wide, and steps and stairs absolutely perfectly regular. This isn’t just a gentlemen’s agreement—we have intentionally designed our world in full knowledge of our common human physical abilities and constraints, and have trained ourselves with these expectations. The less we need to think about navigating our cumbersome, fleshy bodies through the physical world, the more we can focus on more interesting pursuits.
This is never more true than in action games. Whether gamers are playing against a clock, a gnarly missile-wielding octopus robomonster, or simply for a shedload of cash, the last enemy they want to face is the environment itself. Being disarmed of the ability to win because of poor or confusing level design is one of the most deadly sins, and never more so than in competitive matches. The audience at Katowice was there to be entertained by the conflict between rival clans, not between a player and a clumsy staircase.
This problem is amplified by the way we interface with 3D games. To the game engine, each player is little more than a virtual block of solid concrete clumsily leaping through the world at the behest of the player mashing forward, back, left, and right inputs. Modern games may show images of animated, humanoid forms running through rich, beautiful, shapely, often blood-splattered environments, but this is only rich tapestry draped over the cold-hearted 3D physics engines at their core.
As level designers our job is to hide this divide and encourage competitive gamers to be screaming excitedly at each other, rather than angrily at us. In Dust, I had made everything extremely simple to navigate—at worst, players might have to navigate a couple of steps, or perhaps a stack of crates. I wanted players to spend their time engaging with each other rather than running a tedious assault course for the 48,331st time. While I never expected Dust to be popular—or good, even—deep down I didn’t want it to become a tiring faff. I ensured that players could get from A to B as easily and seamlessly as possible so they were able focus on the core game of Counter-Strike instead. Selfishly, I also didn’t want to find myself screaming at myself for cumbersome design choices.
This seems like common sense. Of course you make the world big and wide enough to fit the player, just like modern shops and offices are big enough to let us walk around on our feet rather than crawling on our knees, knocking ceiling tiles out with our bowed heads, or having to avoid pit traps and getting snagged on spikes. Of course you provide enough room for everyone. But in the early days of level design, this wasn’t obvious—the temptation to explore every feature of 3D engines meant that maps were frequently over-designed and excessively intricate to the extent that navigation was infuriating. The best maps employed restraint. Their authors kept things simple, and clean. I wanted to create a location that was not just accessible, but believable and convincing, and so I made Dust as uncomplicated as I could reasonably get away with.
Designing for video games raises tricky issues regarding space and scale. As much as we model game environments on our real-world environments, we must do so within the constraints of our technology, and adjust the game world accordingly. Game worlds often have an exaggerated scale, with ceilings higher and doors wider to account for the fact that players view the world through a TV or monitor that offers barely half of the 120-degree or more peripheral vision that we enjoy in real-life. Try this: Point your arms out straight in front of you, then bend them 90 degrees at the elbow so your hands are pointing as if you’re about to do a pull-up. The area framed between your forearms represents the narrow field of view offered by video games—everything outside is chopped off by the edges of the screen. By widening and deepening our game environments, we get to bring in just a bit more of the cropped world into the visible frame. While real-world design is for humans, our virtual worlds have to be built for our virtual facsimiles.
Dust made use of a few simple tricks to help players find their way around. One of the simplest was adding clearly marked roads and sidewalks. We typically think of these as convenient flat surfaces meant to ease our passage from point A to B, but they also aid navigation and provide senses of direction. These features—by their sheer presence—provide clues and hints to our environment. They tempt us to follow them. Like the Wizard of Oz, Dust used a simple network of wide, gold-brown roads and patterned walls to entice players in the right direction, no matter how lost or how thin their field of vision.
Moreover, the wide and open arrangement of roads in Dust ensured that no matter what tactic or route each player elected to take, there would be ample room to accommodate them. It was only through replicating my school corridors that I had learned the reasoning and value of these allowances.
The two bomb locations in the Dust maps had also benefited from my attempts to copy my school. I had learned that focus points should be distinct, recognisable, and reasonably isolated. But they also need to be accessible. When done well, they are easy to approach but hard to control—not too dissimilar from playgrounds and school halls, as any experienced teacher will tell you.
In team-based games like Counter-Strike, maps typically have entrances and exits arranged in such a way that no matter which team you’re on, it’s very hard to hold an area indefinitely. There’ll always be a way for your opponents to get in. But the winner of any battle will be determined more by skill and luck than by properties of the map. Dust 2 wouldn’t have lasted long if the terrorist team could plant the bomb and sit in a corner with their sights trained on the only way in. Every good defensive position must have a weakness, or a strong offensive position—which in turn must also have a weakness. Designing competitive multiplayer levels is in many ways a game of rock-paper-scissors. Knowing the rules of the game, you design a counter for every advantage, and a counter for each of those, such that, in the end, you can be reasonably sure that neither team has an absolute advantage over the other as they roam from one end of the map to another hunting for victory.
To help players get around and stay focused on the game, Dust and Dust 2 share much in common with the broad strokes of town and city design. The simpler Dust ended up with a hub design anchored by a long central corridor with direct connections to both bomb spots. It’s a design common for both residential and business complexes, where people can use the central hub as a landmark to gain direction and orient themselves. In the game, it’s this central corridor that often becomes the initial focus, with the controlling team then able to dictate how the match progresses. My original—simple and learnable—layout meant that Dust became the perfect playground to introduce new players to Counter-Strike, letting them focus on the core mechanics of the game rather than getting bogged down by navigating the intricacies of the environment.
In contrast, Dust 2 adopts a layout closer to a grid plan—albeit a grid of only three by three. Again, players can use the very center of the grid to navigate their way to the bomb locations, but they are permitted more opportunities to shift between locations. It’s this freedom of movement that allows competitive players more room for tactical maneuvering, proving a big draw for professional matches. Yet, at the same time, the simple design remains easy for players to learn and master.
Large towns and cities have historically benefited from these same designs, which are intended to help inhabitants and visitors alike easily navigate between points of interest. Lowering the cognitive load required to get from A to B not only reduces traffic, but leads to increased enjoyment of the environment.
Monsters
So if level design is little more than building a world akin to our in 3D space, then why don’t we see established architects and town planners selling their designs to game studios? It’s because—at the core—we share deeply conflicting values. Level designers, I’m afraid to reveal, are monsters.
City planners and architects spend years designing spaces to help people live and work with the minimum of expense, effort, and time. They work against tight budgets, uneven plots, slivers of safety margins, unpredictable supply restrictions, and inane local laws in the pursuit of functional works of art that let people live and work in comfort and simplicity.
Level designers, however, start out with a mathematically exact, empty void free of imperfections and restrictions, and an infinite number of tools and supplies. Then we intentionally create flawed worlds, sub-par buildings, and dangerous constructs. We go out of our way to make it harder and trickier than it could be for people to achieve the goals that we set, all for the sake of blood-soaked entertainment.
John Carmack, the brains behind Doom’s 3D engine, famously said, “Story in a game is like a story in a porn movie. It’s expected to be there, but it’s not that important.” This is especially true to multiplayer games. Counter-Strike has no story and Dust has no story. Any suggestion of a relationship to our own world exists purely as a framework on which to hang an enticing central premise. Dust thrived by being the thinnest-possible veil over the addictive core gameplay of Counter-Strike. It imposes itself as little as possible on the players, adopting a visual framework inspired by African and Turkish architecture, but without sharing any of the same goals or values.
The maps of modern competitive games aren’t designed to be real so much as they are designed to be fair and fun, and to achieve these ambitions we have to do terrible, brutal things—from blocking up perfectly serviceable doors and routes to erecting walls in which to contain our lab rats. As much as level designers try to emulate the design of the real-world, we spend just as much time trying to work around the limitations those rules impose. We strive to create a world the player can understand and relate to, but then we add elements—not for their function—but because their presence is indicative and convincing of a deeper, richer world than the one we have actually constructed. None of the windows or doors in Dust open, but their presence hints at the structure of a 3D world beyond what the player can see on their 2D screen, serving as minor landmarks that help them get around.
While players and designers value consistency, we also absolutely treasure surprise and unpredictability when it comes to multiplayer games. A good map must allow for players to express individuality, rather than leading them down the same path each time. City design is the opposite—a congestion-free city relies on predictive models of human behavior to funnel people around, and there’s an expectation that everyone will behave more or less the same every day. In designing levels for single-player games that players might only play one or two times, that’s perfect; but for multiplayer games, where each level will be played hundreds if not thousands of times, it becomes incredibly tedious to play or watch. We have to design with chaos and choice in mind, even though this conflicts with our goal of producing a fair, balanced environment.
Good building design insists that hallways are free from obstacles, protrusions, and doors that open inward (lest someone from the connecting room open it in your path at the exact moment you’re running past to get to your next class, for example). Anyone entering these highways is supposed to be sure of which direction they’ll need to walk in and where that’ll take them. Hallways are the motorways of our structures—simple, streamlined, and easy to use.
But in Counter-Strike, it’s rare to find a long, empty hallway. The fact is, empty hallways don’t play well—put a player at either end and within seconds the one that survives will be the one with the fastest reflexes or (more likely) the bigger, badder gun. It doesn’t make for exciting or interesting gameplay, and it’s not what the eight thousand spectators in Katowice came to see (not forgetting the hundreds of thousands of spectators who watched online). Players dislike such situations, for these situations can’t call upon the skill and tactics they’ve so carefully honed over time.
So we add elements like crates, pipes, and destruction to provide cover and an opportunity for either player to advance tactically. Dark alcoves offer excellent opportunities for players to hide and surprise the enemy team, and you’ll often find that we’ve placed them in just the right place to keep a game interesting, even if they serve no other purpose.
. . . and that’s why the familiar locations we see in computer games have more crates per capita than any place on Earth. For their flexibility, level designers have a disturbing love of crates.
Perhaps more disturbing is how we take the good, sane, and sensible principles of structure design and use them to influence players, encouraging them to go one way rather than another, hoping to tempt them into the tricks and traps we’ve set up. Yet, at the same time, we want them to break the rules, to surprise each other, and even us. We do everything we can to make a playground for fair competition, but one that also exploits expectations inherited from the real world.
In essence, good level design is not achieved by knowing the ground rules of good space and building design, but in perverting them. Level design is often knowingly the antithesis of best practices, with maps carefully dressed up such that we hope you don’t notice our tricks and traps, yet still have a blast gunning down your friends and foes. We pick and choose from the world we know to create one we don’t, curating aspects of our environment, shuffling them around, and reseating them to give a sense of the familiar in pursuit of something fun.
The four-day tournament in Katowice concluded with a gripping final between Virtus.pro and Ninjas In Pyjamas playing in two other maps that knew the same tunes and tricks as Dust 2. To the casual spectator, the finalists may have appeared to be fighting through a random assortment of alleyways and hidden backstreets, exploiting a battery of obstacles, and finding cover between seemingly misplaced barrels and crates. The arenas—presented as beautiful, picturesque locales—were postcard perfect. In truth, they had been tuned and tailored to the extreme. They were skilfully designed to pit the wits and reflexes of gamers against each other.
These e-athletes weren’t lab rats in a synthetic maze anymore. They were experienced masters of their environment, defeaters of the cruel rules and artificial restrictions that had been designed to control and entertain. They had moved beyond containment, using tricks and tactics of their own making—beyond those even imagined by the designers. Just like the world we take for granted lets us live, build, and thrive, these maps let them focus on what they do best—play, adapt, and win.
David Johnston is a London-based game developer and level designer. Apart from his world famous Dust (de_dust) and Dust 2 (de_dust2), he has worked as a level designer on games such as Counter-Strike: Condition Zero and Brink.