So You Want to Learn 3D Level Design

Maybe you are a student of game development and want to specialize in level design. Maybe you are a game developer already, but want to learn some new skills. Or maybe you’re a hobbyist and want to build strange new worlds. Whatever drew you to 3D level design, you may not know where to start.

If your goal is to make an entire game, or if your focus is learning the tools used by professional level designers, you should look into learning the Unreal Engine or Unity.

However, if your goal is to learn the conceptual skills of level design, to go beyond learning the tools and practice design, then you are in the right place! Practicing design requires a context, and with level design that means building levels for a game. If you build 3D environments without that context, you will learn the tools, but you won’t learn design.

With this goal in mind, this tutorial series covers the basics of designing a level for Quake.


E1M5 loaded in the Trenchbroom level editor

Why not Half-Life 2, which has stronger thematic variety and more assets to draw from? Or why not Counter-Strike: Global Offensive, which has an active mapping community? Why Quake?

  1. Quake has modern tools that are easy to learn and use. This lets us minimize our time learning the tools so we can focus on design fundamentals instead.
  2. Quake has a tight focus to its gameplay. There is a small range of weapons, enemies, and environments, and these constraints keep the design context simple. These constraints also let us avoid the distracting features of modern game engines.
  3. Quake has an active mapping community that will playtest, offer feedback, and help you check your design ideas against reality.

But, learning 3D level design through Quake comes with some downsides: the beneficial constraints can be an expressive limitation; we don’t have preview lighting in the editor, which makes lighting a little tricky; and also, some of the construction paradigms are specific to Quake’s tools.

Once again, if your goal is to make an entire game, or to learn the tools of professional level design, you should look into tutorials for the Unreal Engine or Unity. My goal with the tutorial series here is to help you practice design skills with minimal technical obstacles.  And whatever your goals are, I hope these tutorials will offer some new insights!

In Part 1, I walk through the setup for the game and the tools. In Part 2, I cover the map compiler and the basics of level editing. From there, we’ll explore the design patterns for Quake, and how to create memorable levels.

Making Your First Map in Quake (Part 1)

Release Notes for Toybox

Last weekend I released my Quake mod “Toybox”, which offers two new maps of open-ended exploration with an emphasis on joyful movement. You can download the mod from dropbox or quaketastic.

If you don’t have Quake but are still curious about Toybox, here’s my video playthrough:

Or keep reading for my design notes below!


Map Pack or Mod?

When I released Toybox, I wasn’t sure whether to call it a map pack or a mod. The enemies and items are all still Quake in their function and presentation. The code changes—all part of progs_dump by dumptruck_ds—are there to simplify level creation, but most of the gameplay in Toybox is possible without modified code.

The most striking modification in Toybox is how the levels look. Quake has never been so bright and colorful before. However, I did not change the color palette. These colors were always available to modders, but seldom used. Even before we add colored fog and lighting, which are features of modern sourceports, Quake has what it needs to be bright and colorful instead of dark and brown.


Quake’s color palette

My level design for Toybox is the other modification. In Quake, most levels offer a sequence of obstacles to overcome. Some of these obstacles, like locked doors and keys, serve as goals to guide the player and structure their progression through the level. Other obstacles, like traps and combat encounters, are the heart of the experience; these obstacles ask players to react with skill or die.

Instead of this structure, I built my Toybox levels for a nonlinear progression. Both levels have two keys, which together open the exit, but players can choose the order they seek them. There are multiple methods to acquire most of the keys, and there are other objectives in the form of runes and secrets for the players to pursue. My hope with the open layout and objectives is for players to explore at their own pace.

Importantly, I didn’t use secret tags. In normal Quake maps, players can track how many secrets there are to find. But with my Toybox levels, some secrets aren’t possible on the first playthrough. Specifically, both levels have secrets that require runes from the other level. If I tagged these areas as secrets, and players saw they had some number of secrets left to find, I worried that my players would spend hours looking everywhere in the map instead of moving on. By not tagging any secrets, I hope my players exit the level when they are satisfied with their exploration.

I also included a tourist mode instead of the traditional easy mode. This option lets players explore the world without any monsters to fight, and lets players focus on exploration, level interactivity, and the joy of Quake’s movement. Because the keys in each level aren’t tied to monsters or combat encounters, it is possible to complete both levels in this mode.

Playgrounds and Living Worlds

Some of my design goals for Toybox started with “Recursion”, a map I released in August 2018. With that map, there were four goals for the player to complete in any order. Completing each goal would reset the level to the start, and reset the enemies and items with it. One of my inspirations for this was Super Mario 64, which sends the player back to the hub after completing each goal; this reset may have been for technical reasons at the time, but it was also a way of keeping the world alive and changing across multiple objectives. For Recursion, this reset meant that players lose the territory they claimed after each reset. The player’s role was no longer to conquer the dungeon, but to pass through a world that would persist without them.

With Toybox, I tried this a different way. When the player enters the level, there are monsters roaming, and the player can kill them all before proceeding with any of the goals. However, many of the monsters aren’t a threat until the player wanders nearby, so there is no urgency to clear the level. With this as the baseline, each key the player finds adds a new horde of monsters to the level. Some of these reinforcements pose immediate threats to the player, but others are there to reclaim territory and offer new challenges as the player continues to explore.

Both Toybox and “Recursion” are imperfect attempts to create a world that players can’t consume. Ideally, the monsters would respawn on their own. Or instead of dying, the monsters would freeze or stun for several seconds

For “Recursion” and Toybox, I drew inspiration from the design of playgrounds. In my work, I want to create persistent spaces for non-prescriptive play. But to achieve this goal, I’ll need to mod Quake in ways that change its identity.

Toybox as Subversion

Most of the Quake community focuses on the craft. There are over two decades of work refining techniques specific to Quake and its tools. Most of this work exists within Quake’s language of dungeons, monsters, and gore. A few works, like czg’s Honey, push to the fringe of traditional aesthetics and expand the language of what a Quake map can be. But, compared to the Doom mod scene, Quake remains conservative. Even in the Quake-inspired and Quake-derived retro-shooter renaissance of Dusk, Prodeus, and Wrath, it’s still monsters, dungeons, and gore. (So much focus on gore, you’d think that’s all Quake is missing…)

My hope with Toybox is to nudge this conservatism. Within the Quake modding community, I want to encourage others to explore outside the established techniques and craft. I also want to signal to a wider community of developers that there is room here to make strange and beautiful worlds. I want people to know that Quake can be more than its dungeons and monsters and gore. Of course, Quake is excellent at those things too; this isn’t a one-or-the-other situation.

Through the simplicity of Quake’s modding tools, I see the potential for a new generation of developers to learn the skills of 3D level design. But for this to be possible, we have to create space where different forms of expression are welcome. We have to show outsiders a reason to join. One of my hopes with Toybox is to start making that space so more experimental and personal work can follow.

There is more I want to explore in the ideas around Toybox. I want to do more with the design language of playgrounds and welcoming environments. To do this within Quake, I have to modify the code more deeply and push my work away from this ambiguity between mod and map pack. With its own identity, my project will lose its power as a subversion of Quake.

So, while there is more I want to do, I felt it was important to release Toybox in its current form. In this release, version 1, Toybox occupies a happy space as both mod and map pack, as both Quake and something all its own.

Special Thanks

To close out this post, there are some folks I want to thank:

First, thanks to dumptruck_ds for his progs_dump mod framework, and for all of his support in the Quake community. Without his work, the Toybox levels would have been a pain to create, and a few of their features would have been impossible. It is also thanks to his tutorials that I started making maps for Quake around this time last year.

Second, I want to thank Benoit Stordeur (aka Bal). He was the first playtester for the mod, and his feedback and enthusiasm motivated me to continue with these ideas. Without his encouragement, I might not have finished Toybox at all.

I also want to thank the level design community beyond the mod scene, especially those of you who see what I’m doing here and say “huh!” with legitimate curiosity. Even if you don’t get a chance to play this mod, I hope you’ve found this post useful for your own work.

Lastly, thank you for reading!

Notes on Welcoming Spaces in Games

Last week I posed a question to my friends in level design and environment art: how do we create welcoming spaces in our games?

I received many great answers on shape theory and lighting, answers on specific objects for the feelings they evoke, and answers on mechanics to create a welcoming space. However, the variety of answers exposed a flaw in my question. What does “welcoming” mean?

When I posed the question, I framed “welcoming” as the opposite of feeling apprehensive or wary. Level design and environment art have many techniques to make the player afraid to move forward. We also have shorthand and tricks that have become clichés over the decades: the blood trails and graffiti warnings, the flickering lights and tilted halls. Within this context, I asked what techniques we have to achieve the opposite effect of apprehension.


System Shock 2 (1999)

I received two types of answers to my question. The first type considered “welcoming” as a feeling about a space. That is, if we imagine a room with no door or world outside, what makes the player feel welcome in that room? The second type considered “welcoming” as a feeling of contrast while moving from one state toward another, better state. That is, if we imagine a building in a world, what makes the building more welcoming than the world around it?

My hope, in turning this conversation into a post, is to clarify the question and highlight some ideas for us to explore further. Eventually, I hope we will techniques as robust—and even clichéd—as our toolkit for creating apprehension and dread in our games.

Welcoming as Contrast and Movement

To start, imagine again that abstract building in an abstract world.

But let’s be specific. Imagine a spooky forest on a cold night. Imagine a cozy cabin on a hill in a clearing. Wandering the forest, we hear wolves howling. Right as we think we’ve lost our way, we see a path leading up to the cabin. We see the warm light of a fireplace illuminating the windows. We see smoke curling gently from the chimney. We approach, still unsure if we should ask for shelter or continue on our way, but we hear music and laughter within. This is a welcoming feeling, to be drawn or attracted toward an appealing space.

Now instead of a forest, let’s imagine a nuclear wasteland. We hear the flesh-hungry mutants howling. Our gas mask filter is worn out, and we could use some clean water and food. Instead of a cozy cabin, there’s a rusty metal shack. It shows no signs of life, or death, which is a good thing when there are bandits around. A dusty window will let us peek inside to be sure it’s safe before we enter. Maybe there will be supplies to salvage? At the very least, the shack means a moment of shelter on our journey.

Is this still welcoming? We should be cautious approaching the shack because this is a hostile environment, but the shack itself is neutral or better than the world around it. We are not repelled, and the possibility of supplies is attractive.

For both of these examples, the buildings offer the possibility of satisfying our needs for survival. There is the possibility of food, water, and rest. But even if we can’t satisfy these needs, the building itself offers shelter from a hostile environment. Both buildings are more positive than the world around them, which makes them attractive. But are both welcoming in the same way?


Near Death (2016)

Before we go too far, let’s look at a specific example from games. In Near Death (2016), the player is trapped at an Antarctica research station in a blizzard and has to find a way to survive. There are no monsters or enemies, only the cold and the wind. The buildings also aren’t enough on their own to warm the player. For that, the player uses a portable heater and a limited supply of kerosene. In some buildings, the windows are blown out and need patching before the player can use the space to recover and warm up.

Because of these mechanics in Near Death, venturing out to each new building is a gamble. There may be supplies, like more kerosene. Or there may be blown-out windows that demand supplies to repair, adding an element of attrition. But, as the player moves through the research station, the known and safe territory expands. Backtracking into a room the player knows to be safe produces that welcoming feeling.


To turn these examples into a more abstract case, we have something like the diagram above. The player is in a hostile world, sees a safe building, and moves toward it. Arising from the environment’s context and from the movement toward a positive space, the player feels welcome.

The exact cause and effect is not clear here. Does the player experience the welcoming feeling before deciding to move toward the building, or after? Are the feeling and action simultaneous, or is one in response to the other? These are difficult questions. My hunch, borrowing from appraisal theory, is that the emotion follows initial stimuli and cognitive appraisal of the environment context.

If we change the environment context, or change the direction the player is moving, the player experiences different emotions.


This diagram is a fuzzy way for me to condense several of my subjective experiences, so treat it with caution! The emotional labels here are imprecise. The importance of this diagram is to isolate the welcoming feeling from related experiences, which deserve their own investigation.

Specifically, moving toward a positive state may not be enough on its own to make the player feel welcome. There is some aspect of entering a building or an enclosure that contributes to the welcoming feeling. This may be an association with safety and refuge, but I’m not sure if this real-world association is strong enough to override when a game’s mechanics make interiors more dangerous than exteriors.

Valence Theory

What about more complex examples? What if we remove the threshold between building and world? For example, what if instead of a building we have an opening in the forest? There is no door to open, but standing in the clearing will feel different from standing among the trees. Or what if there are two different cabins to choose from, or a whole village?

To handle these more complex cases, we can borrow from what Robert Yang dubbed “valence theory”, which was itself adapted from Randy Smith’s GDC 2006 model of “Level Building for Stealth Gameplay”. An important feature of this model is that the states within the environment are fuzzy and imprecise. The states have gradients of intensity and overlapping boundaries.


A node layout from Aubrey Serr’s GDC 2019 talk

Aubrey Serr’s GDC 2019 talk “Designing Radically Nonlinear Singleplayer Levels” also adapted Randy Smith’s model, but with a strong emphasis on the cloudy quality of each state within the level. Serr’s diagrams add objectives (indicated by stars) to explain why players move through negative states instead of staying where they are safe.

With these more complex networks and fields of polarized space, what does the player feel? Is the nuance of a welcoming feeling lost to the less specific feeling of tension and release? There is more for us to investigate here.

What Makes the Player Move?

So far I’ve described the welcoming feeling as a product of moving from a negative state toward a positive one where there is some element of enclosure and safety. But what attracts the players in the first place and motivates them to move?

With my example of the cabin in the forest and the shack in the wasteland, I assumed human needs and interactions. We need food, shelter, and water. If there are people in the cabin, we could talk to them. If there are supplies in the shack, we could salvage them. With video games, we can’t count on any of these real-world assumptions.

For example, the Doom games invert our expectations. For us humans, hell is bad, and Earth is (comparatively) good. But for doomguy, who only wants to rip and tear, Earth is boring. Doomguy runs toward hell.


Beyond Doom’s simple inversion, we can make our valence model a proper mess by looking at states that simultaneously attract and repel the player.

Let’s imagine the forest scenario again, but now as a horror game. Instead of a cabin, we have a haunted house. Instead of the warm light of a fireplace, we see some eerie glow of mysterious origin. Instead of laughter and music, we hear chains and electric crackling from somewhere below.


The forest in this imagined horror game is not pleasant, but also not hostile. Looking around, we see a fence some distance into the trees, and if we checked, we’d find that we are unable to cross it. We also have no way to interact with the forest, and there is nothing to do but walk in circles or approach the house. Let’s say this imaginary game also has no mechanics for hunger or thirst, and we don’t need to worry about hypothermia outdoors. Within this context, why does the player leave the safety of the forest to face the danger of the haunted house? Something attracts the player at the same time the threat of danger repels.

In Miriam Bellard’s GDC 2019 talk on “Environmental Design as Spatial Cinematography”, she described several ways that environments attract players and influence their movement. First, there are mysteries that draw the player to seek information. A towering landmark in the distance may not be enough to draw the player, but if there is a mystery around the landmark, then the player has a reason to investigate. Second, Bellard describes multiple ways that object affordances motivate the player to move through a space. To oversimplify a complex topic: people tend to move toward areas with greater affordance opportunities. Or, in game terms, players tend to move toward interactivity. (Watch Bellard’s talk for the more nuanced explanation.)

To return to our horror game example, mystery and affordance draw the player at the same time that the aesthetics, and the dangers they suggest, repel us. This tension of both wanting and not wanting to proceed is at the heart of horror games. The player feels a dissonance of wanting to know the mystery but not wanting to face the danger on the way.


The start of Gone Home (2013), which hooks the player with mystery.

But the welcoming feeling is in opposition to mystery. In my example of the cozy cabin in the woods, we may hesitate because of the unknown, and we need reassurances to overcome that uncertainty and feel welcome. Or in my example of the shack in the nuclear wasteland, we would be far more apprehensive if there was no window to peek through before we step inside.

Without mystery as a tool, what remains from Bellard’s toolbox is the attraction of affordances. Inside the cozy cabin, or inside the shack, we know there are things to do. But if the game offers no ways to interact in these spaces, then there is no attraction; the cabin and shack are no longer positive spaces for us to be drawn toward and feel welcome.

However, there are forms of play not represented within game systems and mechanics that can still attract the player. Non-interactive spaces can still serve role playing, imaginative play, and social interaction.


“Lon Lon Ranch” from Ocarina of Time as depicted on

For example, in Legend of Zelda: Ocarina of Time (1998), there is “Lon Lon Ranch” where the player has several optional activities to earn valuable items throughout the game. Once the player completes these activities, the game no longer offers incentives to spend time there. The space becomes minimally interactive. There are still things to do, like chasing the chickens, but the game offers no reason to interact beyond intrinsic enjoyment. Yet, when I played Ocarina of Time, I would return to “Lon Lon Ranch” between the game’s dark and scary dungeons. This was a welcoming space.

Beyond the ideas of affordances and mysteries, there is more we can explore:

  • Are players drawn toward aesthetic beauty, and in what contexts?
  • What about safety and comfort?
  • Or community and a sense of belonging?

Before moving on to the idea of welcoming as a state, here’s a summary of what we have so far:

  • The welcoming feeling arises while moving toward a positive state, often enclosed, where there is the promise of safety and interactivity.
  • The player moves toward affordances, mysteries, and rewards. The player’s movement may not align with the welcoming feeling at all!
  • If we want welcoming spaces, we need to create the reasons and opportunities for our players to enjoy them.

Welcome as a State

Back at the start I said how the responses to my question fit two types. Now that we’ve explored the welcoming feeling as contrast and motion, let’s explore the welcoming feeling as a state.

Thankfully, a 2017 Project Horseshoe report about “Coziness in Games” did most of the work already! As part of the section on cozy aesthetics, the authors suggest literally welcoming the player into a space:

When the player is explicitly positioned as a welcomed entity, this gives them the freedom and safety to express themselves. This welcome does not imply responsibility or pressure on them as a hero … but rather welcomes them as a person, to join whatever activities are available, or to be alone, as they wish. Bartenders often greet newcomers with a welcome, whether the tavern is digital or physical, to encourage a longer and more leisurely visit.

The other patterns for cozy aesthetics are about safety, protection, and abundance. A welcome environment is one where players are supported and feel like they belong.

Welcoming as Dominant Mode

Most of these cozy spaces described in the Project Horseshoe report are small rooms that fit within one screen. If we expand these spaces, we may lose the cozy aesthetic, but how far can we push them while the player still feels welcome?

In some levels, the welcoming feeling of safety is the dominant mood. There may be pockets of risk or danger, but they are isolated with clear boundaries.


“Bob-omb Battlefield” as depicted on

For example, the first main level of Super Mario 64 (1996) is sunny and verdant. There are friendly NPCs to greet the player at the start, and the first real obstacle the player encounters is the Chain Chomp. This enemy is literally leashed to a pole and can’t escape. A red coin on top of the pole encourages the player to toy with this boundary. Within this safe context, danger is made a tool of play. The other enemies in the level also have leash-like behavior where, if the player runs far enough away, the enemies will give up the chase. But the boundaries here are more subtle than the Chain Chomp’s.


The Chain Chomp enemy in “Bob-omb Battlefield”,  taken from the Mario Wiki

If we return to our valence theory from earlier, this first level of Super Mario 64 looks a bit like the diagram below. For each of the level’s risks, the players can retreat to safety and try again when they are ready.


“Bob-omb Battlefield” as depicted on, with our valence diagrams on top.

Mette Pødenphant Andersen’s GDC 2019 talk “Hitman Levels as Social Spaces” also highlights the idea of spaces where the player is welcome. In the level “Sapienza” from Hitman (2016), a welcoming public space surrounds a walled villa. Between these public and private spaces, there are clear boundaries so players know when they are trespassing and toying with the level’s risks, much like the player’s experience with the Chain Chomp in Super Mario 64. Other spaces have softer boundaries, but if the player can’t manage the risks, then the large public space allows the player to retreat and find another way.


The social spaces in “Sapienza” from Mette Pødenphant Andersen’s GDC 2019 talk.

How far can we expand these welcoming levels? Can we make the whole game welcoming?

For Super Mario 64 and many of the 3D platformers like it, the welcoming levels are only there at the start. As players progress, the levels ask more of the players’ abilities, and the level aesthetics turn dangerous to match theses mechanical demands. As a consequence of these dangers, the early levels like “Bob-omb Battlefield” become even more welcoming when the player returns. Or for Hitman, “Sapienza” becomes a vacation from harder levels like “Colorado” where there are no public spaces. In both games, these welcoming levels become a genuine reprieve because of the harder levels ahead.

Ideas to Explore Further

Phew! I hope these early notes have highlighted some techniques and ideas to explore. There are also more specific ideas that I didn’t cover. In particular, I am left with a few questions that I hope to investigate in the future:

  • We feel welcome when visiting a friend’s house, but we don’t use “welcome” to describe the feeling we have in our own house. Why?
  • At what point does welcome becomes mundane and boring?
  • Is the welcoming feeling holistic, where one element out of place disrupt it? Does this make shorthand techniques impossible?
  • Bonfires! How much of the attraction to bonfires is a real-world association that we bring to games, and how much is an in-game association with mechanics?
  • “Welcoming” as an act, or ritual that we perform for our guests. What games explore this?

Thanks for taking the time to read! If you explore these ideas in your own work, or are aware of more games I should study, I hope you’ll let me know :)


At Home in the Level Design Community

Several weeks ago, I wrote that I had stopped playing games. I wrote that I didn’t know when I would return. Now, after a week at the Game Developers Conference, I am playing games again. As an explanation, there are many thoughts I want to unpack. For those of you who make games, I hope these thoughts are useful for your own journey.


Since moving to Canada in August, I’ve sought community. Living in a new country alone is difficult, and in my attempts to balance my life with my work, I’ve needed a life, a community.

At my work here I may be the youngest member of the level design team. My coworkers have families and houses. For them, the studio is their second, or third community. And for me, these coworkers serve as mentors for how to live a life in balance.

But because of our differences, my work is also a second community. To seek a life outside of work, I’ve connected with the arts scene through poetry readings and theater. This is a slow process, but half a year since arriving in Canada, I feel I’m at the threshold to a community here.

Then, two weeks ago, the Game Developers Conference happened.

In November, when I was still failing to find a community, the Level Design Workshop at GDC announced an opening for talk submissions. I pitched my talk on applying playground to multiplayer level design. In the submission form, they asked why I wanted to present. I said that I care about the level design community, that this community has given me mentors and friends over the years, and that my talk is one small way to try to give back. Later that month, the Level Design Workshop accepted my pitch, and I set to writing.

As GDC approached, my life and my efforts to find community shifted. I stopped playing games. I fell out of touch with the mod scene and the Quake community on which my talk relied for many of its examples. I felt a reckoning in myself from the debts of overwork in games, and I felt cynical.

With each email GDC sent about biz-dev networking opportunities and talks about optimizing monetization strategies, I felt my cynicism deepen. One cold email invited me to talk over coffee about “competitive intelligence within the mobile app landscape.” Considering I don’t work in mobile games, and that “competitive intelligence” means nothing to me, I have to assume this email was one of thousands weaseling for business opportunities.

To be clear, I’m not a capitalist. Each day’s news that the planet is warming faster than expected pushes my socialist tendencies toward anarcho-communism. To me, the game industry’s eagerness for profit and growth has no rational place on a dying planet. In these critical years before our carbon-emissions pass the point of no return, it seems unlikely that game development is the best use of my life. But all of my skills are in level design, though this may be a sunken-cost fallacy.

So, receiving these emails from GDC, I felt a dissonance within me. On one side, I was committed to present my talk and give something back to the level design community, as I had promised. On the other side, I felt a repulsion to the business, the tech-obsession, and the history-rewriting carnival of the games industry. I don’t believe games are the highest medium for art, nor that games could, or should replace books in the classroom. I don’t believe in inevitable progress, or that the newest technology is best. But these beliefs are at the heart of GDC. It is a business conference, after all, and there is a self-righteousness in these values that I don’t accept.

If you are steeped in the culture of games, or if you are a student eager to work in the industry, you may not understand this. I may sound entitled, ungrateful. Here I had the privilege to attend the biggest conference in games and present a talk. There are many who could have made more of this opportunity, appreciated it more. And if my cynicism had persisted and I had phoned-in my talk, you’d be right to fetch pitchforks and light torches.

In that first day at GDC, I felt like an imposter, not for want of skill or knowledge as a craftworker, but an imposter like an atheist attending church.

Then the night before the Level Design Workshop and my talk, we all met for dinner. As a bunch of level designers in a room are wont to do, we talked shop, we talked the stresses of the job, and we turned nostalgic toward the mod scenes and the tools where we had learned our craft.

When I mentioned Unreal Engine 2.5, I got a laugh of recognition. When I said the Quake mod scene is still active and thriving, there were smiles of delight. As dinner carried on, I observed I was the most animated I’d been in months. I wondered if this was my response to the crowds of San Francisco, my travel-stress turned raw energy. But our conversations were exciting. Now, more than a week back from GDC, I am still excited by the ideas we shared.

As a group of level designers meeting for dinner, we left our company loyalties at the door. We met like freelancers gathered in a moment of peace. Most of us there were employed on long-term contracts, and some were with studios where they have worked for years. But in games, we are all freelancers in the long term. Our industry is turbulent, layoffs are frequent, and we are all trying to find our place. Later that night, one of the veteran designers told me he never expected he would be in games this long, yet there he was, still finding ways to do meaningful work.

In the conversations at dinner and in the days that followed, I felt at home in this community of level designers. This is a specific subset of the industry, not just level designers, but the small group of us who feel absurdly compelled to share our knowledge, give talks, and explore the craft. We are an academic bunch, not businessmen. Many of us come from the anti-commercial folk-traditions of modding. We are the historians of our craft because who else will care, if not us?

When I call this a community, I mean we are a group where I feel I belong. For me, community describes a feedback loop where my work has value as part of an ongoing conversation, and community offers other voices from whom I can learn. There is the sense of safety in a strong community, to explore new ideas. Above all, this community is unified by the curiosity to explore the possibilities of our craft.

For some developers, GDC is about renewing a faith in games with fresh inspiration. This was my experience when I presented a talk in 2018. But this year, this was not about faith. This year, the Level Design Workshop within GDC reminded me of the community where I already belong. After months of seeking a community in Canada, so slowly, GDC reminded me of the community I already have.

This troubles me in many ways. The level design community has its problems. We are skewed toward white men and fail to support or remember the work of women and people of color. We are skewed toward English-speakers who can afford to spend a week in San Francisco. As a subset of the industry, and a subset of GDC, we share in the larger biases of game development. These biases limit the conversations our community can have, and they limit our potential as a craft.

I am also troubled because this is a once-a-year community. We carry on our conversations by shouting around the loud dinner tables and bars of San Francisco. This leaves me wanting more: to talk longer, more clearly, more often. But GDC ends and we return to our studios across the globe until the next year. If the community where I belong is there among level designers, I’m horrified how we are splintered over distance and time. If this is my community, then my community is broken.

To say this troubles me is an understatement. I feel something closer to horror, or despair, like discovering I am on an island alone. I feel reminded that we are all freelancers in the long term, that making games takes a kind of homelessness. In an industry this turbulent, the only lasting community may be this broken thing we have together.

And, frankly, I’m not sure what to make of this, or what to do.

Returning to Canada from GDC, my thoughts could be summarized as “well, shit.” If the community where I belong is a collection of international level designers, then our once-a-year reunions may be as good as it gets. The arts scene in Canada I’ve spent months pursuing may only ever be my second community.

However, I realize for us level designers separated across the globe, part of our conversation is through our work. It may be a year till we talk again in person, but we are still in conversation so long as we listen.

When I wrote a few weeks ago that I had stopped playing games, I said I didn’t know when I would return. Now that I’m back in Canada from a week at GDC, I find myself returning, slowly at first, to the study of games. For now, this is the way I know to keep in conversation with the community where I belong.

Thanks for reading,

Playgrounds & Level Design

At GDC this year, I presented a talk on applying playgrounds to multiplayer level design. Because this was a short talk, I focused on the application to level geometry and the consequences on multiplayer systems.


If you’ve followed my blog for the last few years, you’ll know my interest in playgrounds is not new. They are a metaphor I’ve used several times as an example of how digital games should adapt to their to players’ needs. What follows is my framework for these ideas in multiplayer level design. There are still many areas to explore in this space, some of which I outline along the way, but I hope this introduction to the playground metaphor helps you in your own explorations.

The Basics
Playgrounds are persistent community spaces for non-prescriptive play.

A playground is not an obstacle course to complete, a puzzle to solve, or a story to hear. There are aspects of obstacle courses and stories, but they are not consumed through player interaction; they persist. A good playground facilitates many obstacle courses, many stories, and many forms of play.

Playgrounds are places to be, not levels to beat.

Playgrounds are Multiplayer
Playgrounds are fundamentally multiplayer spaces. They support a wide range of player needs, and they scale along game mode, player count, and player trust.

Raph Koster and a team with google explored the idea of a Trust Spectrum in games. The idea is that different multiplayer activities call for different amounts of trust, which can create barriers for an audience. If a game supports basic parallel play alongside intense, cooperative raids, there is not a path for new players to build the trust they need for the harder activities.

Playgrounds also support a trust spectrum with a range of play, from solo and parallel activities up to group competitive and cooperative play. As multiplayer spaces, playgrounds also follow some design patterns for building friendship. The activities encourage proximity and reciprocity between players.


Here, for example, are several activities as part of a playground. Players can enjoy some activities alone, or in parallel. Other activities require multiple players, some with turn-taking, others with players working together, which means reciprocal loops that build trust. These activities also don’t lock players in for long; a player can stop in the middle of an activity, or at the end of turn to opt out of play.

When I describe playgrounds as community spaces, I also mean more than the play structures. The community space includes the surrounding park, the walking paths, and benches. Playgrounds include spaces for parents to sit and talk while their children play. This is an area I want to explore further in video games, but what could serve the function of a park bench in our multiplayer game systems?

Playgrounds, Not Sandboxes
There is some overlap between playgrounds and sandboxes worth clarifying. For me, a sandbox is a neutral space where players bring toys. In contrast, a playground is the toy. Sandboxes are also used up through the act of play; often a player will need to start a new game for the sandbox to reset. The main overlap between these terms is that both allow non-prescriptive play. That is, sandbox and playground design both let the players do what they want instead of forcing one way to play.

Sandboxes best support solo and parallel play, but not rich cooperative or competitive experiences. This may be a consequence of genre convention more than a necessity. When I say that sandboxes are neutral spaces where players bring toys, I mean the individual player’s upgrade trees, loadouts, and equipment. Because these are not shared, or passed between players in turns, the game systems encourage each player to specialize in a way that may conflict with deeper cooperative play.

For example, in Saints Row IV there are a finite number of activities and collectibles in the world, and the act of play depletes this world of its potential. While exploring the world cooperatively, there are few activities that require both players. As we level up, we can each choose to spend our skill points differently. If I focus on enhancing my character’s movement abilities, while you focus on combat, then the systems are pushing us away from cooperation. You won’t be able to catch up with me while I scale the buildings, and I won’t be able to fight alongside you when you attack harder enemies. Again, the emphasis of the sandbox design is on the toys—skill upgrades in this case—that each of us collect and bring to the neutral space, instead of the space being a persistent toy for both of us to share.

My focus with this post is on level geometry, but the way we shape our levels affects multiplayer systems. If we play Deathmatch on a King of the Hill map, the game may still play like King of the Hill. This is because level geometry and game systems work together to create the play experience.

The Formula
Now let’s dig into some examples:


Here is a basic playground from my neighborhood. If you live in North America, you are probably familiar with playgrounds like this. As playgrounds go, this one is simple and unremarkable. Importantly, we can’t convert this structure directly to our video game levels, and we wouldn’t want to, but there are lesson we can take from its design.


The basic unit of a playground is a module, a self contained chunk.

At the center of the module is a platform elevated off the ground. This playground is for young kids, so remember this platform is a meaningful vantage point for a five year old. This means it functions as a goal for that player to attain.

In order to reach the platform (the goal) the player must pass a skill check. Here that skill check is a staircase. Importantly, a good skill check is not a binary pass or fail. The player can fail and trip on the stairs, but most use falls into a range of skill: for the three year old, slow; for the five year old, OK; for the ten year old, maybe two steps at a time?

Remember the multiplayer context for these skill checks. If I can take the stairs two at a time in a game of tag, and another player can’t, then there is a tactic I can exploit around the playground.

On reaching the platform, there are routes back down that function as rewards. Here they take the form of slides, which are also opportunities for subversive play for the skilled player who will risk running up the slide. Subversive play is whenever the player thinks they are using a constraint of the design against its intent.  We can design with some subversion in mind, but the flexible form of playgrounds should allow for more subversive play than we can predict.

Against the bars at the back of the platform, there is a steering wheel. This functions as a setpiece, which landmarks and differentiates modules from each other. In playgrounds, a good setpiece can also spark imagination and encourage roleplaying. (For now, the aesthetic and thematic solutions to the playground formula is a problem to save for another time.)


Now that we have a formula, we can create more complex forms. This playground is more interesting than the first example because it connects multiple modules into a complex structure.


Here we have a hierarchy, or tree. This makes a fun playground, but there is still a problem in the lack of cycles or loops. If we played tag on this structure, the only ways to go from one end to the other are by backtracking or by running on the ground. But with our abstract language of modules, we can imagine variants that do connect as cycles. With these formulas, we can now imagine infinite, sprawling playgrounds if we want!

But our imaginary playgrounds do hit a constraint of safety. Real playgrounds are safe environments for failure, where the design limits risk. Playgrounds also limit the time to recover from a failure. If I lose my grip while crossing the monkey bars, I lose a few seconds of progress before I can try again. In games, if the player fails a skill check and cannot recover in a short amount of time, then the level lacks the safety of a playground. If the player reloads a save state, we’ve designed something where recovery is too difficult. This may limit how big a playground can be and what systems we can use.

Applying the Formula to Level Design
For most of my experiments with playground level design, I’ve used Quake. The most direct application of playgrounds requires a robust set of skills, which Quake has. A staircase in Quake is not a meaningful skill check, but once we add a monster at the top, there is some skill involved.

Another benefit of testing these ideas in Quake is that all levels support singleplayer, co-op, and competitive play. Once we adapt the level geometry, we’re close to my definition of playgrounds.

There are also many ways to create timing-based skill checks around moving obstacles, traps, and dynamic cover. For a game two decades old, there is a lot we can do!


Quake also has advanced movement skills:






Bunny hopping lets players maintain their momentum, and circle jumping (curving their aim while strafing in air) lets players accelerate and curve around obstacles.






Jumping at the peak of a ramp gives players more height than normal. Jumping a little too early or a little too late will still give some bonus to the jump height, which makes ramp jumps one of the easier movement skills to use.






Rocket jumping grants even more height, but it comes at the cost of some health. We also usually reserve the rocket launcher as a rare reward item.

These advanced skills create opportunities to move through the level quickly and take shortcuts, but if we made them necessary skill checks, the level would be inaccessible for most of our players. That is, these advanced skills should be fast routes and opportunities for subversive play, not the only routes.

Putting it all Together


Putting these ideas together, we get a basic playground. This level plays in a different way than a standard Quake level, and I want my players to recognize this the moment they enter the map so they won’t rely on existing habits.


If we break the structure into its modules and their connecting skill checks, this should look familiar. With different skills involved, this same formula could describe a real playground.


Pushing these ideas further, adding more skill checks and rewards in the form of enemies and items, we get a pretty cool structure! Importantly, while there are objectives for the players to complete, there are many ways to approach them, many ways to add new goals, and many ways to play.


Speedrunning is one example of players creating a new goal on top of the intended design. In the video below, one of my playtesters routed a speedrun. Normally a first playthrough of the map takes 10 minutes, but here it takes 27 seconds. This speedrun also demonstrates the range of subversive play possible within this design.




(If you aren’t familiar with Quake, the speedrun starts by grabbing a secret Ring of Shadows, which makes the player invisible for 30 seconds. Normally, this is enough time to get an edge on combat or explore in safety, but not enough time to complete the level. Importantly, if the goal of speedrunning is set by players outside the game, then they can decide if they want to use the item. Speedrunners create their own categories and rules.)

For comparison, and a more thorough exploration of the level, here is my playthrough:


Other Applications for Playgrounds?
At this point you may be thinking “but Andrew, I’m not making that kind of game! How does this help me?” And probably what I’ve described is very different from what you’re making. But if you work with 3D level design and you follow the needs of camera, character, and controls, then playgrounds are still useful.

If you’re involved in prototyping or pre-production, you need metrics test maps. Normally, a metrics test map is the 3D equivalent of a meter stick to know how high, how far, how fast, and so on. But these maps can’t tell you whether your metrics are fun. So instead, metrics playgrounds!


Here’s an example of a metric playground I built to find the fundamentals of a game early in production. We had a tight third-person camera and a robust move set with double-jumping, mantling, sprinting, and crouching. In a standard metrics test map, there is no way of knowing if your character moves too fast or jumps too high. Aside from verifying your code works, distances only matter in the context of play. The purpose of my metrics playground was to provide that context.

As we continued building the game, we kept the metric playground up to date with our mechanics. We had gunplay, so there were ways to spawn enemies to shoot and others to evade. In this way, the metric playground let us test the whole game in one spot and quickly tune for fun.

Would our final levels look or play anything like these playgrounds? Probably not, and that’s OK! The purpose of the metrics playground was to find our fundamentals.

I am not suggesting playgrounds are a magic solution to all of our problems. Rather, playgrounds are another area of design we can study and—when appropriate to our goals—apply.


To me, playgrounds are an exciting metaphor for level design because of how they facilitate a breadth of play that we, in video games, seldom allow. They are level geometry doing the work of system design, and they show a way forward for levels that adapt to their players’ needs.

Thanks for reading,