Why build a blockout?

A blockout (also blockmesh or graybox) is a 3D rough draft level built with simple 3D shapes, but without any details or polished art assets.

The goal is to prototype, test, and adjust the foundational shapes of the level.

In the image below, notice the differences between the blockout version and the final shipped version. A shape might start as a gray block -- then after months of playtesting and art passing, the block becomes a stack of barrels... or maybe gets deleted entirely.

Blockouts support experimentation.

It is "cheap" to delete or rebuild some rough blockout geometry, but throwing away finalized art passed work is "expensive" and wasteful.

When you are not confident about the level structure, it is best to keep it "cheap" until it is ready to become "expensive."

You can't playtest a design document or a layout sketch, but you can playtest a blockout and evaluate its flow, balance, encounter design, and metrics. This is the design phase when you finally start to discover whether the ideas will work or not. The blockout is just the beginning.

"... If we consider that [Neon White] shipped with over 100 levels, the team made “double” that amount in scrapped levels, and [the level] Smackdown alone had more than 50 changes made to it, we’re faced with thousands of tweaks and redesigns throughout the entire game. [...] Three years later, Smackdown is complete."

-- from How A Neon White Level Is Made by Blake Hester, Game Informer #350

Key concepts

When blocking out, consider these aspects:

• Massing is the general sense of volume and weight conveyed by the shapes.

  • Is this structure thick / heavy, or thin / light? What kind of place is this?

  • Landscapes need special consideration.

  • Composition is currently over-emphasized in level design culture today.

• Metrics are the general scale, dimensions, and proportions of the level.

  • Is this area too big or small? Can the player fit in this room?

  • examples of useful measurements: Doom metrics, Quake metrics

• Wayfinding is the player's navigation process for learning the map structure.

  • How to help the player find the critical path / level exit? Does the player feel too lost?

• Playtesting is when you run an experiment to see if the level meets its design goals.

  • Can most players complete the level? Do the encounters work? Is it balanced?

  • Playtesting is really important. This is the whole point of making a blockout.

Construction methods

There are five common 3D level construction methods in games:

• Primitives: arrange simple basic shapes like cubes and boxes.

• Brushes / modeling: construct 3D shapes in the level editor.

• Modular kit: connect pre-made pieces together, like Lego.

• Sculpting: "paint" organic 3D shapes, useful for landscapes.

• Splines: create math curves that procedurally generate geometry

Primitives

Primitives are simple 3D shapes, like the default gray cubes and boxes built into a game editor.

Move / rotate / scale the shapes. Arrange them together, stack them on top of each other, make more complex groups of shapes.

This is a surprisingly powerful blockout method. You can always use this method in any 3D engine or tool, but you'll hit limitations with blocking out anything beyond simple boxy buildings.

Brushes / modeling

Brushes are simple low polygon 3D shapes modeled in the level editor.

This was the main level construction method across the industry from 1990s-2000s.

We strongly recommend this method because it offers the most control and handmade feel, but unfortunately most modern game engines often have poor brush modeling tools. It can be tricky to figure out a workflow here.

Modular kit

Snap together pre-made pieces of architecture from a kit of modular parts.

However, if the kit is badly designed or poorly configured, it'll be hard to use. If you're a beginner, you probably shouldn't try to make your own kit -- just download one instead.

For more on planning, measuring, and constructing modular kits, see Modular kit design.

For links to download free modular prototyping kits, see Resources.

Sculpting

Most modern engines have a sculpting tool that lets you paint and deform a flat plane.

We use this only for making big smooth shapes like terrain and landscapes; it's bad for hard surfaces like buildings.

But if landscapes are central to a project, there's no way to avoid sculpting.

For more on sculpting and designing terrain, see Landscape.

Splines

Splines are invisible curves that generate geometry along the path.

These curves are also non-destructive; you can adjust a spline at any point and regenerate its geometry.

A spline is ideal for blocking out curvy linear forms like roads or rivers, where it can also automatically deform the surrounding landscape.

It's also possible to generate non-linear areas and buildings with splines too, but this type of workflow isn't common and usually requires custom tooling.

How to blockout

Here's a general process. If you're a beginner, or you're at the start of a new project, try to do everything. You will gradually personalize your own process and style.

1. Sketch layout

2. Add ground plane, scale figures, walls

3. Playtest

4. Diverge, iterate, and playtest again

5. Repeat step 4 until done

1. Sketch layout

It can be difficult to blockout without any plan, especially for beginners. Even a simple scribble can help a lot.

Look at the layout sketches below. The left sketch emphasizes the scale of each space, the middle sketch focuses on the relationships between the areas, while the right sketch is a more detailed floor plan drawing. Any of these sketches can work, any of them can help you plan a blockout.

Just draw something!

2. Add ground plane, scale figures, and walls

Create a ground plane (or a big flat cube) near the center of the 3D space (0, 0, 0).

Use a light-colored gridded prototyping texture so that you can visually estimate sizes and scales. This floor object will help "ground" the rest of the level geometry, providing a horizon line and context for everything else to rest upon.

Add a player-humanoid scale figure to help establish scale.

If you don't have any humanoid assets, create a roughly human-shaped block to serve as a placeholder. The scale figure helps establish consistency and clearance for the player.

For more on common hallway / wall sizes and humanoid dimensions, see Metrics.

For links to free prototyping and blockout textures, see Resources.

Build a wall segment that's approximately 150-200% as tall as the scale figure.

If you're working in a modern game engine, sizes don't have to be exact -- the sense of proportion is most important to establish here.

If you are blocking out with a modular kit, just place a standard wall module next to the figure.

Texture the wall with a different color from the ground plane. Color and brightness provide important context for understanding what the space is made of, and how these various floor and wall planes relate to each other. If you are afraid of colors affecting your blockout too much, then at least use textures with different shades of gray.

Copy & paste wall segments to build up more space until you have at least one room or hallway. Some tools feature special "duplicate" or "clone" commands: in Hammer, hold Shift and drag the brush; in UE4, hold Ctrl+Shift and drag the wall object.

Intersections are OK. Messy is OK.

To make a doorway or window, just leave a gap between wall segments, and fill it in later. Entryways and hallways should be at least twice as wide as the scale figure.

3. Playtest

Do a simple self-playtest: walk around the space, within the game engine, with full player gravity, collision, and speed.

The purpose of the blockout is to experiment. To verify the results of the experiment, you must test the prototype. Do NOT just fly around in the editor view, that won't help you imagine the player experience.

• Massing. Do the level's shapes make sense? Is it confusing to look at?

• Metrics. Does the player fit through every door and passage? Do the stairs / ramps work? Can the player jump where they're supposed to jump?

• Flow. Does movement feel slow or fast, smooth or complex? What type of movement is ideal for your intended design goals?

4. Diverge and iterate

Diverge from original plans and respond to the playtest.

99% of the time, your blockout will not survive a playtest. Rooms will feel awkward, doorways too narrow, the layout too confusing. But this is exactly why we blockout!

Want to rebuild huge sections of your blockout? Go for it. Want to delete an entire room? Do it. Need to split a courtyard into two smaller areas? You have permission.

Keep an open mind about what your playtests and walkarounds are telling you.

Continue adding new scale figures, walls, and floors.

Then playtest again.

5. Continue iterating

Keep an open mind and continue this cycle of modification + playtesting. Continue developing the blockout. Build, then walk around in it, then modify, then playtest again... and repeat. This design process is called iteration, because we are making new iterations that build off of the strengths of the previous version.

As you gradually build more and more of the level, it may surprise you how much it changes over time. Let the map surprise you.

Common blockout problems

• (most common) My blockout feels too small / too big.

  • Use more scale figures during construction.

  • Playtest more often and catch scaling problems sooner.

  • Too big? Delete unnecessary rooms and compress what's left.

  • Too small? Expand outermost walls, then expand the center to fill up the new space.

  • Or delete everything and rebuild. It's just a blockout.

• I have anxiety and stare at the empty level editor screen without making anything.

  • Sketch a layout drawing, and then try to follow that plan.

  • Just blockout one simple room. You can always delete it later. The point is to get rid of the blank page. In drawing, this is "activating the canvas"; in writing, this is a "shitty first draft."

  • Do something else and return to the blockout later.

Example blockouts

"Castle" for Dirty Bomb (Splash Damage)

Splash Damage level designer Anthony "MassE" Massey designed the map Castle for the competitive multiplayer team shooter Dirty Bomb. Massey took inspiration from twisting medieval streets and typology of the real-life Tower of London, and conceptualized the initial layout drawing (above, left). But note how the resulting blockout (above, center) differs from the layout with fewer curves, more 90 degree angles, and chunkier areas. The layout was just an initial guide. Because this is a multiplayer shooter map, the design team playtested the blockout with a focus on movement and combat metrics:

“The scale of maps is always the hardest aspect to get right. It’s important to plot your main paths and measure these distances; we’re looking for anything between 8 and 12 seconds from spawning to an objective. [...] Anything longer than that and players get frustrated when they respawn and have to run back. On the other hand, if the time is shorter it risks the map feeling too small for 8v8 matches, and can lead to chaotic gameplay. [...] This stage is crucial to map development, but our team operated by a simple rule of thumb; if it feels long, it’s too long."

"Considering [player classes] is the next step. We had to look at combat ranges to allows our entire arsenal to function, [...] while also ensuring variety in combat spaces to allow [different abilities] to be viable (like [an airstrike ability]). In real terms, this means considering the ratio between outdoor and indoor areas, ensuring [characters with outdoor-focused abilities] were viable."

-- from "Blocktober: Dirty Bomb" Splash Damage blog post

"World's Edge" for Apex Legends (Respawn)

Respawn level designer Rodney Reece built the blockout for "World's Edge", a large competitive multiplayer map for first person battle royale game Apex Legends. Notice how the map changed drastically from blockout (left) to art pass (right) in not only theme but also layout and composition.

"Originally the idea was the map was snowy. But art wanted to bring green into the map, and in one of the meetings, Robert Taube suggested what if the Epicenter (then called Frozen Explosion) caused the snow.

A key to being a good designer is being flexible to new ideas. For instance, the Art team came up with The Dome. In my blockmesh, it was a volcano. They pitched it and I adjusted to incorporate the idea. I had to create a new layout for it, but it made it better.

But other times, certain [points of interest] are fun from the beginning. In those situations, I am more stubborn about what can change. Sorting Factory is a good example of that. It's important to identify what's precious and what's flexible in terms of layout. Because it takes a team!"

-- remarks by Rodney Reece (@RodneyReece7)

Against blockouts?

For some projects, the traditional blockout might be less helpful. The experience may depend less on spatial design and more on the art pass. Blockouts can't validate a concept dependent on art or other assets.

Vertical slice for "Firewatch" (Campo Santo)

For the first person narrative exploration game Firewatch (2016), developer Campo Santo wanted to focus on mechanics like walking and talking; the main appeal of the game concept was looking at art passed scenery and listening to voice acted dialogue. Following typical best practices, they first built a blockout to test the viability of these mechanics. However, the blockout did not help them answer any questions about the player experience because the game pacing was fundamentally a narrative design and environment art issue, not so much a level design issue. The traditional blockout process wasn't working.

According to environment artist Jane Ng's account, it wasn't clear whether Firewatch would "work" as an experience until they skipped the blockout process and instead completed a vertical slice prototype with an art passed environment and near-final dialogue.

Location scouting for "Untitled Goose Game" (House House)

For the top-down third person stealth puzzle game Untitled Goose Game, developer House House sought to create an authentic-feeling British village in 3D. Level designer Jake Strasser had never been in the UK, so blocked out the initial level with occasional use of photo reference. However, this blockout-first approach required Strasser to fill-in gaps from his imagination -- and because he didn't have a British imagination, these gaps often felt implausible or inauthentic in subtle ways, counter to their intent. The traditional blockout process wasn't working.

So instead, Strasser tried a research-first "location scouting" approach. He went on an exhaustive virtual tour of various UK villages, taking lots of photos from Google Street View. After recreating several village streets in-game, Strasser finally pin-pointed what type of structure suited the game's needs, and settled on an unusual side street layout based on Pump Street in the village of Orford. Here, blockout was less about building a space, and more about discovering a space.

To review

A blockout is a simple but playable "rough draft" level with minimal detail. We keep it simple and blocky so that we can easily change it.

You can build a blockout with primitives, brushes, modular kit, sculpting, or splines. The best technique will depend on your engine and project, or you can combine multiple techniques.

How to blockout:

1. Sketch a layout, even a simple one

2. Add a ground plane, scale figures, and walls.

3. Playtest immediately; does it feel OK to walk around? You need to know now.

4. Diverge from your original plan, iterate and playtest again

5. Repeat step 4.

The most common problem is scale, especially for beginners. Playtesting the blockout in-game is the best way to know if it feels too big or too small.

Blockouts might be less helpful for some projects like single player art-dependent narrative levels.

Now what?

• Review key concepts for blockouts like massing, metrics, wayfinding, and playtesting.

• Once you're built and tested a blockout, do a scripting pass or a lighting pass.

More about blockouts

• GDC 2018: "Invisible Intuition: Blockmesh and Lighting Tips to Guide the Player and Set the Mood" by David Shaver and Robert Yang is probably the most up-to-date industry-standard blockout talk, showing blockout examples from Shaver's work on Naughty Dog games alongside distilled examples prototyped in Unity. However, it's actually more concerned with composition and wayfinding than construction.

• "Quake Mapping Tips: Building Layouts" (5 min) video by Michael Markie starts with a very simple one room blockout in Quake 1, then gradually elaborates on it and makes it more compelling to play. Note the frequent playtesting and design iteration. A great example of an improvised blockout process with little pre-planning.

What is massing?

In architecture, massing is the overall feeling and logic of shape and space.

• Does a room feel large and heavy, or low and hidden, or light and open? Why?

• Compare two rooms. Are they close together or far apart? Why?

• Does one wall seem related to another wall? If so, how and why?

Ideally, the massing supports the building's program, the overall logic and function behind the building's organization.

In the massing diagram above, an architect proposes a new addition to a public library. Each mass has a defined function and audience.

The new entrance hall (orange) is at street level on the right, and the children's area (yellow) is near this entrance so that children don't walk too far or disturb others. The central open space (teal) is a long courtyard that acts as a core community space and primary circulation with the existing library (gold), highlighting the unusual cylindrical form as people approach from the entrance.

The symmetry of the added library wings (green) matches the symmetry of the existing library. The street level retail space (dark gray) is near the entrance hall, so that people can conveniently visit both the retail and the library in the same trip. The depository (light gray) needs street level access for loading / unloading shipments, but takes advantage of wide floors with low natural light.

Now apply massing theory to levels. How will players use each area and when?

(TODO: massing diagram example for games)

(TODO: paragraph description of massing diagram example for games)

Massing methods

DIMENSIONAL

The simplest way to adjust a shape is to move, rotate, or scale it in the 3D level editor. This is best with basic shapes like cubes and other 3D primitives. If you try to stretch or squeeze detailed meshes and modular kits, it will probably look weird.

ADDITIVE

Make more complicated shapes by combining simpler smaller shapes. This is the most common massing method in level design and real world architecture.

Articulation: when shapes feel separate and distinct, we say they are articulated. Articulation can make a building feel smaller when the parts feel appropriately sized for people, built at a human scale.

SUBTRACTIVE

Carving a simpler shape into a more complex shape. This is helpful for modernist futuristic architecture, or for carving organic-feeling natural rock forms.

Beveling / chamfering / boolean operations are key techniques in 3D modeling, but most level design tools today lack clipping tools or subtractive CSG support.

Continuity: in contrast to additive massing, subtractive massing often feels more continuous. Continuity unifies the entire form as one single monolithic shape. Big monoliths won't feel human-sized.

Hierarchy

Some masses are more important than others. A bigger mass with a more unusual shape will feel more important than surrounding masses. If this big mass stays continuous, it can interrupt other masses and take priority over the other shapes.

In the photo below of Seinäjoki City Theater designed by Alvar Aalto and Elissa Aalto, the main body of the building feels wide and low, somewhat articulated into overlapping blocks. But what's that big brown mass erupting out behind? It is a large tall monolithic shape with unusual angles. It is a contrast with the other boxy shapes, and the color / material is different too.

In the floor plan above of the Seinäjoki City Theater, notice how the building consists of 4 main masses: three linear grids for a restaurant, backstage, and offices, orbiting around the central concert hall space (see emphasis in upper-right). For a theater, obviously the concert hall is the most important part of the program, so that's why this unusual shape defines the rest of the space.

(TODO: video game example)

Readability

Readability is how the level suggests its layout and organization using its appearance.

A map with high readability arranges relatively simple shapes into clear distinct groups that the player can easily measure and memorize.

In contrast, a map with low readability has a confusing structure and shape, with fragmented massing that feels like camouflage. But low readability isn't necessarily bad because less readability can still aid your design goals. Broken massing complicates a shape and divides it into smaller shapes, and sometimes that's what you want.

In the photo above of 580 De Haro St in San Francisco, the large residential block of townhouses has been broken into a dozen different overlapping shapes, colors, and materials, which suggests many different young unique quirky families live here.

If the architect had kept it as one big continuous monolithic shape, then it would feel much bigger and less "fun", possibly diminishing the character of the surrounding neighborhood and their property values. (Or you could argue it feels like a desperate attempt to dress up some condos and mask the true nature of gentrification sweeping the city.)

Imagine 580 De Haro as a video game level. Imagine the layout inside. How does this massing affect the player's wayfinding strategies?

• Articulation: each part feels like a separate townhouse for a separate household. The townhouses probably do not connect to each other, and each interior space probably only has one front door / street level entrance each.

• Continuity: yet all these townhouses were built together and form one big overall sense of place. Although they are different, they also feel related. Maybe they have similar layouts, circulation patterns, and surface finishes inside? It wouldn't be surprising.

• Culture: in real-life, have you ever seen modern townhouses like this before? If your audience is familiar with this type of building and if your level uses realism, then they may expect the layout inside to feel plausible.

Proportion

Sometimes you don't even need a grid to establish proportion / a center of gravity

Thickness and Weight

Thickness matters. Thin floor vs thick floor, thin cover vs thick cover, narrow lip vs wide beam

in general, avoid thin floors / walls, prone to tunneling and clipping issues

thickness as gameplay: bullet penetration and wallbanging (siege, csgo, blops4)

Corners

slice the pie, sharp vs rounded (link to Balance?)... sightlines

Simple corners

A simple corner doesn't draw attention to itself. It exists as part of a larger shape or volume. The shape and texture doesn't vary, the only thing that makes it pop is the way the light hits it.

Line of sight blocker if 90 degrees or less

If it's a subtle bevel, don't do it in blockout

Rounded corners

corners with obtuse angles (> 90 degrees), unified shape, make the player flow / move faster around it

Opens up line of sight, less effective as cover

Great for cars and vehicles, smooth circulation

Complex corners

Detailed reinforced corners emphasize the corner as a distinct edge, and transform the corner edge line into an object. The corner is less about the the overall volume, and now it feels like more of a place or a thing to use, with its deviation from the walls and its own distinct silhouette.

More complex corners are good for:

• framing a certain view, perspective, or vista

• making the boundary / enclosure feel more solid, permanent, and grounded

• emphasizing a corner as a cover object / line of sight blocker, where the player should dwell -- but in this case, don't over-detail it with small gaps, keep it fairly chunky and solid-feeling

Open corners

Break up the edge with an opening, to emphasize how the walls are planes that enclose an interior. If lit head-on with thick walls, the open corner can feel very subtle and barely-there, in case you don't want players to notice it at first glance.

Sneaky corners? a trap?

very modern, very rare in traditional masonry and carpentry

When enclosed with glass: fancy airy expensive

Avoid shape psychology

Shape psychology is the theory that certain shapes convey universal ideas and influence behavior.

This book argues strongly that shape psychology is 99% bullshit. Abstract geometric shapes do not make all humans feel the same thing nor communicate the same ideas. Humans form mental associations with shapes based on a complex combination of genetic predisposition, personal experience, and culture.

Even if it wasn't bullshit, shape psychology would not necessarily transfer to a video game context. People often behave one way in the real world, but behave differently in a video game. Player behavior depends more on game patterns, mechanics, available information, cultural framing, and roleplaying persona, rather than whether a shape is round or square.

To design for consistent player behavior in levels, you should playtest instead. Do the work, instead of wishing for a magic theory to brainwash players.

For more on debunking this brain poison, see Shape and color psychology.

What is landscape design?

Honestly there's not much existing theory or documented knowledge about level design and landscapes. Environment artists tend to work at a small scale with limited portfolio piece scenes, while level designers focus on buildings. When you look for tutorials or guides on this topic, almost everything will focus on engine-specific sculpting tools and geological simulation tools -- and nothing about the actual act of designing the terrain. Our existing design theory has basically neglected the landscape, which feels foolish when so much modern shooter, open world, and multiplayer design happens in landscapes.

Unfortunately, real world landscape architecture provides only limited insight for level designers. The landscape architect must account for surrounding ecology, climate, hydrology, soil chemistry, archeological mitigation, land rights, building codes... these are interesting factors to consider for worldbuilding, but definitely out of scope when you're just trying to blockout some basic level geometry.

So in this section, we want to focus on design and function. We define landscape design in video game levels as the purposeful shaping of topography to aid desired player flow and experience goals.

Anatomy of a mountain

Landscapes are made of rock, dirt, and sand, gradually worn down by wind or water (erosion).

When the tectonic plates along the Earth's crust interact, they often push rock upward (tectonic uplift) and form a variety of mountains. When warm air rises and pushes clouds toward the top of a mountain, the higher altitude and lower air pressure causes the clouds to rain. That rain water flows back down the mountain, forming rivers and valleys. That rain stays on the windward side of the mountain range, while the other downwind side (leeward) loses moisture to the hotter air and thus stays relatively dry and arid (rain shadow).

On taller mountain ranges, there is often a prominent tree line that marks where trees can no longer grow because it is too dry, cold, or windy. Above the tree line, the mountain rock is clearly exposed with occasional bushes / plants, or heavy snowpack in the winter.

To design a plausible naturalistic landscape, consider how various type(s) of rock, erosion, and climate will shape the terrain. Then consider how the local inhabitants responded to that terrain.

• Water flows downward and curves around obstacles. That means rivers curve a lot, which also means mountains and valleys curve a lot.

  • ... so roads and paths should curve around as well, since humans must follow terrain. Straight flat highways and roads are relatively modern, and imply a powerful industrial society that can devote (or waste?) resources to grading roads.

• Vegetation holds dirt in place. Trees with deep roots are especially good at resisting erosion and blocking wind. Forests act as windbreaks, which means there can be different biomes on the windward / leeward side of a forest.

• Elevation changes affect wind and water patterns. If your level features mountains, should the biome be the same on both the windward / leeward

Landscape sculpting workflow

People have been sculpting for thousands of years, so there's a lot of theory about how to sculpt, and it's a topic that deserves its own book. Here we're going to focus on the basics of sculpting as they relate to sculpting terrain in a game engine:

1. Layout: draw a simple diagram of the landscape with labels. What are the different biomes and microclimates? Which areas are most important?

2. Blockout: sculpt rough shapes for the landscape, with attention to flow and scale. How big is each valley, hill, mountain, or canyon? Is it high or low, long or short, sharp or smooth?

3. Metrics: measure travel times, heights, and traversable ramps / slopes. Is it clear where the player can go? Do locations feel too close or too far?

4. Art pass: sculpt and paint smaller details, apply set dressing. Does this space have a clear mood? Does it feel lived-in or plausible? "Finished"... or empty?

After each stage, we of course strongly recommend playtesting. When playtesting, check for:

• Areas that feel too big, routes that feel too long or too annoying

• Blocked areas that shoudn't be blocked, unclimbable slopes that look climbable

• Game breakers where the player can get stuck or fall into a crevice

1. Terrain layout

First, draw a layout to plan the landscape. Shade regions according to their theme and local climate. Label and highlight prominent areas. Your landscape layout drawing should be able to answer these questions:

• What are the biggest areas, which areas need to be next to each other?

• What is the theme of each area, and how does it relate to adjacent areas?

• What types of art assets will each area need?

In the example layout drawings below from World of Warcraft, note how the dev team began with a simple bubble diagram to begin solving the spatial relationships. As they spent months / year iterating on the level, the final landmass retained similar organization but resulted in very different sizes and shapes. The "Highmaul" area in the top-left roughly tripled in size, the flood plains halved, and the beach expanded into a harbor. Plan to deviate from your plan.

Also note the clear regional color coding (brown = impassable mountain, yellow = arid, green = temperate, dark green = wet) and how it conveys the overall spatial hierarchy. The top-third is a wet area north of the river, the middle third is a temperate valley, and the bottom third is an arid mountain range.

2. Terrain blockout

Site planning

• Where does the player start and what is the critical path, if any? What's the primary and secondary circulation through each zone, and between zones?

• Does the massing feel small and cozy, or narrow and claustrophobic? What is the spatial identity of each area in the overall composition?

Define the palette of big recurring shapes and objects common in this area (e.g. a house; a tree) and include their placeholders in the blockout.

For big development teams like Blizzard that need to coordinate production across multiple departments, this blockout phase is also a good time to do research and planning with environment artists:

"Working with an environment artist, the level designer will help to guide and define the scope of environment assets needed. These assets include terrain textures, trees, bushes, accent plants, rocks, etc. The range of models and textures needed must address not only the main zone look, but the sub environment types needed to break up the zone, all the while bringing the concept to life while remaining within the capabilities of our game engine. It can be a challenge, and often is.

Take for example the new Nagrand. Not only do we have the environment that you know of as the Nagrand from Outland, but new areas, like a wetlands, and a higher elevation arid region. The visual clash of these disparate environmental themes could be quite jarring if not handled with care. There is a constant conversation between the level designer and the environment artist about shape language, color, diversity, scale, mood, model usage, and ultimately the visual tone of the zone as a whole that keeps the zone development moving in the right direction." -- Ely Cannon, senior level designer on World of Warcraft, from "Artcraft - Level Design part 3"

3. Terrain metrics

Now is a good time to think about balance... shallow slopes vs deep slopes, thin ledges vs thick ledges

Include generous space for landings at the top and bottom of ramps / stairs, or else players might overshoot and jump off a ledge

Is this walkable area readable, are the boundaries clear? Can the player survive a fall from __ height, can they walk up this __ degree slope?

Travel time, where is the nearest fast travel point

4. Art passing

Refine shapes and overall composition and add some details (grass, plants, rocks). Is the silhouette, geology, and erosion pattern plausible or consistent? Work on lighting, mood, and color palette. Does each sub-region feel distinct and readable?

Unity extra terrain tools: https://docs.unity3d.com/Packages/com.unity.terrain-tools@2.0/manual/index.html

Unity Path Paint tool: https://github.com/Roland09/PathPaintTool

You should still be making big changes to the terrain btw, and each time you do, repeat step 2 and 3

Procedurally generated biomes

Handpainting grass and flowers is time-consuming, and if you ever redo your art then you'll just have to paint everything all over again.

The most common method is to define biomes, thematic palettes of foliage and textures that automatically decorate a landscape. Then the terrain system can automatically decorate itself by procedurally generating the details.

Big AAA open world games invest a lot of in tools, scripting, and environment art to automate their landscape art passing workflows like this. It usually isn't practical to do this unless you're making a big project with a team.

Topology

Sculpting

Use "set height" brush to layout floor planes and terraces

Your general sculpting workflow will be: (1) add or subtract mass, (2) smooth, (3) repeat forever

To define a slope, sculpt terraced "steps" and then smooth over. For narrow ledge ramps along the side of a mountain or canyon, a dedicated ramp tool is probably easier.

Don't oversmooth, don't leave everything a blobby melted mess... sharpen and crease, define shapes

For each region, start with 3-4 basic materials and shades.

Paint with consistency. Don't reuse a cliff material as a ground material.

Do the big basic flat sculpt first, then switch to increasingly smaller brushes to smooth over slopes and add minor height variations to the floor plane.

Avoid using erosion brushes for blockout... these can't do level design for you!! These are for detailing your level later in the design process

Topological patterns

Plateaus are raised landforms with flat tops. Terraces are flat surfaces cut from a slope.

Craters are big deep holes in the landscape. Bowls are shallow holes in the landscape, and hollows are smaller bowls. Think of them as negative plateaus which can also be terraced. Descending into these surface depressions can offer a sense of safety, or perhaps a sense of being trapped.

In real-life, terraced bowls make for fantastic naturalistic amphitheaters that focus sound and provide a big comfortable gathering place. Conversely, a small hollow can offer a sense of closeness and intimacy. In games, bowls are most often deployed as artillery craters in battlefields to offer low shallow cover -- which, we suppose, is its own form of closeness and intimacy.

Combining a plateau with a crater produces a unique volcano-like effect that could also double as a sort of shelter or hideout (above, top-left)

Paths

Path design

Axial paths are straight, angular, and efficient, associated with human power and control.

Meandering paths are curved, round, and inefficient, associated with nature and leisure. pg 91

Paths respond to constraints:

• Users. Who made this path, and who is using this path today?

• Topography. Paths usually follow the curve of the landscape or shoreline. Cutting into the landscape or tunneling is a lot of work.

• Vegetation. Paths usually go around large trees or dense vegetation. Clearing land is a lot of work.

Sharply painted paths with hard materials imply maintenance. Disused / occasional paths with soft materials blur and dissolve.

TODO: Breath of the Wild "triangle composition"

TODO: find BotW screenshots that matches both diagrams

Path types

• Ledge paths are open on one side, closed on other. Common for hiking trails on mountains and canyons. Useful for suggesting scenic views, or offering a pleasant contrast between open view vs. closed enclosure.

• Cuttings are enclosed on both sides, usually "cut" through a landform or with retaining walls. It can feel safe and relaxing on a sunny day, or damp and mysterious on a cold dark night.

• Ridge paths are open on both sides and raised up. Common for big open wetland areas, offers a big sweeping exhilarating view if there's enough height.

• Switchbacks or zig zag paths go back and forth with periodic stopping places to control your pace. Not naturally occurring, but often used on steep hiking trails to make ascent (or descent) safer.

Gardening

A landscape is more than the landmass and terrain, it's also about the overall climate, ecology, and vegetation. Landscape architects work heavily with gardening and plantings.

Every video game landscape is a garden, intentionally shaped to facilitate an aesthetic or mood. Video games expect us to suspend disbelief and pretend that diegetically, within the fiction of the game world, that these are untouched wild places. This is a fun lie we reserve for players, but as designers, we should never be taken in by our own lies.

Do research on your desired climate

An ecotone is a boundary where two different ecologies meet. It can be sharp or blurred.

Trees

Trees are basically walls that you can walk through; they enclose spaces through repetition and verticality.

thin enclosure (french garden) vs thick enclosure (a glade in the middle of a forest)

Tree planting: natural or artificial? Artificial: quincunx, formalized forest (pg, 65)

Vary plantings by height

Complex planting: understand humidity, wind, rain shadows, etc

Example landscapes

"Pinar del Mar" Island in Alba, by ustwo games

Alba: A Wildlife Adventure is a 3D open world exploration game about environmentalism and bird watching, set on a Spanish island called Pinar del Mar.

In her post "The Environment art of Alba: a Wildlife Adventure", environment artist Jessie van Aelst wrote a breakdown of her environmental design and development process.

• Pre-production: early on, the team defined four pillars for the entire game -- care for nature, sense of Spain, human impact on natural world, and freedom of childhood.

• Research: they went on virtual tours through Spain via Google Maps, and even went bird watching in real-life. They collected a lot of photo reference and assembled moodboards.

• Blockout: they made an early version of the island with simple geometric shapes and flat colors, allowing them to quickly iterate and test game systems / characters -- once they had a better idea of the game, they later redesigned and rebuilt the island entirely.

For building up the landscape, ustwo developed a custom non-destructive terrain tool workflow in Unity.

For sculpting the main landscape topology, they used a custom Unity editor tool to place splines that procedurally painted onto the internal terrain heightmap at editor-time. This spline-based approach allows for a non-linear non-destructive workflow with smooth flowing shapes that they could fine-tune at will. A designer could also copy-and-paste large landmasses. And because each stroke is a separate game object, that also means multiple people could merge their landscape changes together relatively easily. (note from editor: I'm actually really impressed by this.)

Terrain splines have three brush modes: ridge (for hills and terraces), valley (for rivers and ponds), and level (for flattening / smoothing). The spline's stroke falloff is a user-adjustable AnimationCurve. It also has a built-in terracing curve, as well as a customizable noise modifier.

Each spline pushes / pulls the terrain in a certain way, and the accumulation of all these splines results in a complex landscape.

For texturing, the landscape uses a basic four channel splat (RGBA). Each color (red, green, blue, alpha) corresponds to a texture type -- dirt, clay, grass, and sand. The triplanar terrain shader also applies extra masks for splat transitions, and sets different textures based on world position / world normal.

"For each splat, we had a ruleset. Sand would only appear on certain low heights of the terrain. Dirt would come in as the next layer in the height as well as appear on really flat and even surfaces. Then it would transition into Grass, almost all areas of the game are grass unless the incline was so drastic it would then turn into cliff splat material." "After this pass was done we added some more detail maps to each splat. This means the gradient transitions between splats would now have a bit more detail. These maps really helped with stretching the bandwidth of our limited amount of terrain splats." -- env artist Jessie Van Aelst, from "The Environment art of Alba: a Wildlife Adventure"

They also implemented extra "terrain stamps" overrides, editor-only game objects that could project a specific splat color onto the underlying terrain. This allowed them to add additional details and paint roads / paths, mask out foliage, etc. In the GIF below, note that the last frame consists of hand-placed details, on top of the procedurally generated base.

The team divided the island into several distinct biomes, each with their own plant and animal populations. Artists configured different biome elements for procedurally generated detail scattering / set dressing.

We defined each biome with biome elements. Ex: a pine tree forest biome would include pine trees, pine cones and bush biome elements. The biome elements all have variant slots as well as variables indicating how likely each asset would spawn and what the scale ranges would be.

These biomes would then be placed onto the terrain in a similar fashion to the terrain manipulation. The marker would project cubes onto the terrain to indicate the area where it would generate a biome. The colour of these cubes would represent:

Yellow — area where biome generation happens. Green — an asset will be generated here. Red — an asset could spawn here but won’t because of restrictions

These colours were very useful for us to understand what to expect before generating the terrain. A requirement we came up with was to be able to move around the biome markers and keep the compositions inside intact. For example, a clump of trees that was spawned in just the right composition would spawn in the same composition even if we moved this biome to the other side of the island.

Biomes would generally overlap, so how do we make them work together? In order to overlap or put a biome inside another biome we worked out a system of prioritisation. The marker’s vertical position would therefore decide what would be overwritten.

-- env artist Jessie Van Aelst, from "The Environment art of Alba: a Wildlife Adventure"

Now what?

• Figure out metrics early, especially for landscapes. In most terrain tools, it is difficult to shrink / expand / move different parts of the terrain after the initial sculpt. Moving a mountain is trickier than moving a wall.

• Sculpt a solid blockout first. Resist the urge to sculpt details.

• Landscapes usually won't feel natural until after extensive art passing.

• Narrative-focused landscape projects benefit greatly from worldbuilding.

Further reading

Terrain sculpting workflow

• World of Warcraft: Level Design Panel (Blizzcon 2016) features 30 minutes of level designers Matt Sanders and Ely Cannon sculpting, painting, planting, and set dressing small mountainous island zones in Blizzard's internal map editor for World of Warcraft.

General landscape architecture books

• Form and Fabric in Landscape Architecture by Catherine Dee

• The Fundamentals of Landscape Architecture by Tim Waterman

• Theory in Landscape Architecture: A Reader edited by Simon Swaffield

• The Planting Design Handbook by Nick Robinson

What is composition?

Composition is the overall visual organization of the level. There are two types of composition:

• Spatial composition: the overall big picture 3D arrangement of core masses. We cannot guarantee a specific view, and instead account for a wide variety of viewing angles and positions.

  • This is "composition" as practiced in 3D arts like architecture, interior design, and sculpture.

• Shot composition: how the level looks in the player's 2D camera frame / screen, from a specific angle at a specific position.

  • This is "composition" as practiced in 2D arts like drawing, painting, photography, and film.

A lot of level design tutorials and resources out there claim that specific shot composition is really important. For many reasons, we disagree and we have a different definition of composition. Instead, we argue that spatial composition is much more important than shot composition.

(TODO: image)

Spatial composition

Spatial organization involves building-up a spatial hierarchy, where some parts of the level should feel more important than the other parts. At the top of the hierarchy are key landmarks, focal points, and central areas that contrast meaningfully with their surroundings and anchor the rest of the composition. Some ways of building spatial contrast:

• Height. Big tall towers in crowded spaces, or small short objects in open spaces.

• Density / spread. A wider open space surrounded by smaller narrower structures.

• Orientation. An angled object that breaks from the surrounding grid.

• Shape. Lots of rectangular things nearby? Try putting a round thing there.

In the image above, notice how the central legislative chamber of the Palace of Assembly, Chandigarh by Le Corbusier (bottom-left) uses several types of contrast in its floor plan. It's big, it's a circle in a box, and it's rotated at a skewed angle / axis that breaks from the rest of the grid. So it feels very important and vital to the function of this building! If this were a level, the main setpiece or final boss fight would probably take place in this round room.

The chamber is off-center and almost touches the eastern wall. But it doesn't feel off-center, does it? That's because focal points create new centers within the spatial organization. Imagine a zone of influence emanating from important area; it sort of "pulls" the rest of the composition around it. (See diagrams in bottom-right for how focal points influence how we perceive the "center" of a space.)

And remember, hierarchy depends on local contrast and context. A tall thing only seems special if it is surrounded by short things. A small thing will feel more special if surrounded by bigger things.

Landmark

A landmark or point of interest (POI) is a unique and memorable shape, mass, or location in a level.

TODO: examples

TODO / IMPORTANT: landmarks have to feel relevant and useful !! otherwise they don't function as landmarks!

Actual landmarks (what the player notices during play) vs. fake set dressing landmarks (e.g. random skybox elements in The Last Of Us don't orient players, aren't actually useful)

Saliency

Sightlines

A sightline is a trajectory of empty open space that offers a possible view of another space.

Sightlines are tools that players use when it seems relevant to the current game state and goals. If the player has no reason to look along a sightline, then there is a high probability that they won't use the sightline at all. A sightline is an opportunity / situational tool that the player might ignore, it cannot guarantee behavior.

Even if the player looks in a certain direction, there is no guarantee they will remember or process what they saw. (Imagine reading a boring book or watching a boring movie; your eyes may see the words / images but that doesn't mean you actually read or process it.)

In the diagram above, level designer Sal Garozzo highlights common strategic sightlines on a 2015 version of the competitive multiplayer shooter map de_cache in Counter-Strike: Global Offensive. Each sightline offers a different degree of coverage and information for attacking / defending the bombsite A area.

However, would all these sightlines be readily obvious to a first-time player, at first glance?

No. In fact, recognizing and memorizing these sightlines is part of learning the game / learning the map. Competitive players must actively study these sightlines; it is NOT a natural subconscious obvious way of looking around. The existence of a sightline does not mean the player will use it, nor even know it's there.

For more on sightlines in a multiplayer context, see Balance.

Vista

So how do we convince the player that, on first glance, the sightline is important?

A vista is an exceptionally deep scene composition that offers an overview of the next area, giving the player an opportunity to formulate a strategy. In modern encounter design, it is very common to introduce an arena with a vista beforehand, especially if the arena layout is complicated.

(TODO: talk more about vistas with examples that also emphasize the flow / layout into the vista)

Approach

An approach is a path with a vista.

(TODO: show approach examples)

(TODO: example... Fallingwater?)

Prospect-refuge

Prospect-refuge theory is a widespread belief / design pattern in real world architecture which argues that humans navigate spaces by seeking prospect (vantage points) vs. refuge (safe areas), evoking their innate survival instincts.

We think prospect-refuge is bad theory (and bad science / bad art) with limited applications to level design. Instead we recommend using existing concepts and terminology with less philosophical baggage, like sightline vs. cover.

For more on why we think it's bad theory, see Prospect-refuge.

(TODO: include prospect refuge image)

Shot composition (fallacy)

Shot composition is the 2D arrangement of level geometry relative to the player's camera perspective.

And ok, some of this matters, sometimes. Sightlines, approach, color and lighting, perspective -- all these aspects can matter, sometimes.

But some level designers misinterpret shot composition as level design. They argue that if you arrange everything in your level to make nicely framed views, then players will magically wordlessly know what to do and where to go.

Composition is not mind control. This is such a common misconception in level design theory that we're going to devote a whole section to debunking it. For this reason, we call this the shot composition fallacy: a belief that seems true and useful, but is actually deeply wrong, leading to a worse understanding.

Screenshots aren't levels

First of all, a screenshot is not a level. This may seem obvious, but many level designers seem to have forgotten.

Instead, a screenshot is an image of the game at a specific position, angle, and time. Taking a screenshot is an act of photography.

Do photos capture an objective truth about a place or time? No. A photographer is an artist who chooses a place and time to create an image. Even a dev overview screenshot of a blockout emphasizes and downplays certain elements. A photo is an artistic interpretation that captures only one aspect of an actual person, place, or thing.

No photo can replace the experience of visiting a place or meeting a person, and no screenshot can replace a playtest.

More importantly, there is no guarantee the player will imagine the same framing or composition. The player never plays your screenshot.

Leading lines (are bullshit)

There's currently a trend in level design to analyze screenshots by drawing "leading lines" that seem to point / guide the player's eyes toward a landmark in the distance.

However we argue that leading lines aren't actually effective, and furthermore, leading lines don't even make sense within their own logic.

Leading lines misunderstand human gaze, perception, and information processing. Retina scans and gaze tracking tech show how leading lines and other traditional tenets of 2D visual composition don't "lead the eye" in the ways we think they do. Even if they did, there's no guarantee the player actually processed the visual information.

Literally every hallway you build will seem to converge in the distance. This phenomenon is called foreshortening, creating an illusion of depth in an otherwise flat 2D image. Games with isometric or parallel projection make leading lines impossible. The art history of perspective drawing is interesting, but it's not level design.

Ironically, leading lines mislead level designers into believing that shot composition, environment art, and set dressing are central level design problems instead of metrics, encounter design, or playtesting. Even if leading lines worked (which they don't), you would do it near the end of the level design process. Focus on the fundamental foundation of the level first.

Leading lines routinely fail if you actually bother to playtest. You'll find that there is a 99% chance the player will look somewhere else, and a 1% chance they will look in the intended direction for 0.5 seconds. A screenshot is not a level, no one ever plays a screenshot.

Leading lines are a self-fulfilling prophecy when you move the game camera into a specific place to frame a specific view. Of course the screenshot will follow rules of composition! That's because you composed it like that! And then you drew lines on it! This is photography, not level design. Consult the "Golden Ratio" meme about the absurdity of self-fulfilling composition patterns.

This isn't how people navigate spaces. People use pattern recognition, prior knowledge, and cultural associations to build mental maps in their mind. Wayfinding is a complex active process of moving and using an environment, not just looking at one spot of a level from one angle.

Framing a view within a space is called photography, and it is a skill that humans practice and develop. It is not a natural subconscious process.

Shot composition makes more sense for photography, film, and other static flat 2D media where the author has strong control over the camera, and the audience can pay special attention to the entire image. However, it makes very little sense for interactive environments where the player can move the camera to the wrong place at the wrong time. A video game level is a place, not a painting, photo, or film.

For much better ways of influencing / "guiding" the player, see Wayfinding.

Sightlines vs. leading lines

It may seem like sightlines and leading lines are the same. They both relate to tracing what the player sees from their camera perspective.

However, sightlines and leading lines emphasize very different things:

• Sightline: a trajectory of empty open space that offers a possible view of another space

  • A tool that players use when it seems relevant to the current game state and goals

  • Again, we want to emphasize:

    • Sightlines are made of empty open space

    • Sightlines are situational tools that do not guarantee behavior

• Leading line: fake imagined phenomenon where walls and linear set dressing seems to converge in the distance to subliminally direct the player where to go

  • No one navigates spaces like this

  • No one plays video games like this

  • Don't make levels for no one

To review...

• Spatial composition is the overall organization of 3D masses in the level.

  • Hierarchy is when some things seem more important than others, usually because they contrast in terms of size, shape, spread, or angle.

    • Landmark

    • Saliency

  • Sightlines are lines of empty space that the player can use to see into another area.

    • Vista

    • Approach

• Shot composition is some kinda-bullshit theory that the player's view determines their behavior, rather than the player's actual understanding of the space.

  • Screenshots aren't levels; levels are spaces that players use

  • Leading lines are brain poison, don't use them

  • Don't confuse leading lines with sightlines, which are actually real

Now what?

• Apply this theory to the blockout.

• For more on "guiding the player" with architecture, see Massing and Wayfinding.

Further reading

Our stance against leading lines / shot composition in level design is currently not the norm in the game industry. While we think it's 99% brain poison, other credible designers disagree:

• "Environment Design as Spatial Cinematography: Theory and Practice" by Miriam Bellard, Rockstar North (GDC 2019) goes over Rockstar's approach to filmic environmental composition in Grand Theft Auto 5 levels. Bellard suggests a walkthrough method with rule-of-thirds overlay, and emphasizes affordance / salience model of embodied spaces. She also "myth-busts" leading lines and weenies while simultaneously arguing they are still useful. Overall, this is probably the most "balanced" industry talk on shot composition, but maybe just don't drink too much of the evo psych / shape psychology kool-aid.

• "The Importance of Nothing: Using Negative Space in Level Design" by Jim Brown, Epic Games (GDC 2014)

Prospect-refuge theory is a widespread belief in architecture that humans have an "inborn desire" to seek out prospect (vantage points / lookouts) and refuge (enclosures safe from threats / view).

This book argues that prospect-refuge, as traditionally understood, has limited application to level design. Risk / safety / information in video games often differs greatly from real world architectural assumptions. (And frankly, we're also skeptical whether it's useful for real world architecture.)

At the end of this page, we suggest an alternative application of prospect-refuge theory, using already existing terms and concepts in level design.

(TODO: prospect refuge illustration with big red Xs)

History

In his 1975 book The Experience of Landscape, geographer Jay Appleton combined psychologist Daniel Berlyne's "arousal theory" with early evolutionary psychology to form his prospect-refuge theory. It's worth tracing how this happened.

In the 1950s and 1960s, Berlyne experimented with lab rats and curiosity. He argued that when a lab rat (or a human) sees a new space, they feel "aroused" as they process the unfamiliar environment.

Note that Berlyne does not mean sexual arousal, but instead something more like engagement -- a general state of increased energy expenditure and mental responsiveness. That is, a more complex and novel environment stimulates more arousal / engagement.

However, too much complexity and novelty is negative. In his 1960 book Conflict, Arousal and Curiosity, Berlyne cautions that too much of this environmental arousal can lead to anxiety and uncertainty. When graphed below, it is a "U-shaped" effect:

The basic premise of evolutionary psychology is that humans are just like any other animal with survival instinct. Appleton reasoned that Berlyne's environmental arousal also aided human evolutionary fitness. Increased mental responsiveness would've been useful for a caveman / early human to parse threats and find safe places from predators.

Outside of psychology, no really cared about this until 1988, when architectural historian Grant Hildebrand popularized Appleton's prospect-refuge. Hildebrand wanted to strengthen modernist architecture's claims to universal design and function, and so psychology was an effective way to argue why architect Frank Lloyd Wright's buildings supposedly had intrinsic psychological appeal -- culminating in his 1999 book Origins of Architectural Pleasure. (Maybe these men were indeed alluding to sexual arousal.)

Criticism

There are several ways to critique prospect-refuge theory, including its science and philosophy:

• Berlyne's behaviorism and psychology of aesthetics

  • does studying lab rats' aesthetics actually teach us anything about human aesthetics?

  • can art and aesthetics be studied in this way at all?

• Appleton's addition of evolutionary psychology

  • there is widespread criticism of evolutionary psychology

  • even if evo psych is real, is it relevant to video games?

• The field of psychology is currently in a replication crisis, where famous experiments and theories cannot be validated or reproduced

  • can we replicate the U-shape curve? (2021 study: maybe not)

  • can we replicate prospect-refuge effect? (2016 meta-analysis of 34 other studies: prospect is real, but refuge is not)

At first glance, prospect-refuge theory seems very applicable to games. It is about survival and safety, like many 3D action games. But for the many reasons listed above, we think it is better not to rely on it.

How to use prospect-refuge in level design: maybe don't.

We argue that these same ideas are more useful if we utilize already existing terms and concepts in level design instead. These industry-specific terms have less philosophical baggage and offer more technical specificity, especially for combat games.

• Instead of prospect, use sightline / vista.

• Instead of refuge, use cover / enclosure / base / safe house.

Instead of prospect...

A sightline is a line of empty space where one area can be seen from another area.

A vista is a vantage point with access to many distant sightlines. It is most useful in single player level design to convey progression beats, preview landmarks, and offer layout information -- especially for an arena layout before an encounter begins.

(TODO: link to specialized sightline and vista pages)

Instead of refuge...

The problem with the word "refuge" is that it implies safety. But the video games, the player's sense of safety is complicated, and extends far beyond the level design.

For many action games, threat assessment is a core mechanic / skill that players must develop over time, dependent on the overall combat design.

Survival and death systems also vary greatly between games: some may reward players for death, or even penalize survival.

Safety as game state

A dozen hours into the action RPG Elden Ring (2022), the player enters a "debate parlor" in a magical castle. When most players reach this point, it will feel very dangerous.

• In Elden Ring, the most difficult boss enemies tend to inhabit highly decorated landmark areas. So the unique art assets here differentiate the room from past rooms, it stands out as a special area of high importance.

• The player has reached this point after fighting through 3-5 encounters, and is probably wondering where the next save point is. Past boss arenas featured convenient save points before the fight; this one does not, and pushes players to risk their progress.

(screenshot: debate parlor entry)

Suddenly, a giant glowing wizard wolf begins growling at the player.

• The wolf has a name and a large health bar, indicating it is a dangerous boss enemy.

• The medium-sized room makes the large wolf feel even larger. It will be difficult to dodge the wolf's attacks.

• Furniture and clutter further limit the player's mobility and options. (Until the boss destroys the furniture, opening up more floor space, making the fight a bit easier.)

• Epic dramatic music begins playing.

(screenshot: battle begins proper)

Note all the factors that make the situation feel dangerous: the UI, consistent use of environment art, the past encounters and progression system, even the music.

After defeating the boss, only a quiet "grace" campfire-like save point remains. This landmark boss arena now functions as a safe house, territory won by the player after a difficult battle. The chaotic half-demolished parliament chamber now takes on the mood of a peaceful ruin.

This is the same room, but it undergoes three different state changes before, during, and after the encounter.

• sense of safety depends less on layout / blockout / architecture, and instead more on the wider game systems / encounter design patterns

• game worlds change ownership and function much more frequently than in the real world

• something won't "look safe" if past levels featured traps or ambushes

Thus, "refuge" is a situation / game state, not a formal quality of architecture. No shape or layout is intrinsically safe.

Enclosure

So what word should we use instead of refuge?

We propose enclosure: the degree that an area is surrounded by cover.

Cover is any object that blocks a sightline and/or combat.

TODO: base / safe house

What are metrics?

In level design, metrics are the sense of scale, distance, and measurements across the entire level / game.

All level designers use at least two types of core metrics:

• Player metrics (or physics metrics) are factual numbers measured in-engine like player size, movement speed, maximum jump height, etc.

  • describes game physics, derived from the game engine

  • example: "player size" is a specific fact that can be measured and verified

• Building metrics are suggested floor / wall dimensions and guidelines that you codify for yourself based on your desired experience goals

  • prescribes game feel, decided by designer(s)

  • example: "regular hallway size" is a guideline; what does "regular" mean for each project?

Depending on the project, you might need combat metrics or puzzle metrics too.

But beware of over-reliance on metrics. Numbers and measurements may give the illusion of infallible design laws -- the illusion that following these numbers will give you the perfect level. You must reject this illusion! You cannot measure your way to a good game experience. Metrics are a helpful design tool to help you make decisions, but metrics are not magic.

Scale

Metrics help you establish scale, the general sense of how big the game world feels.

In level design we usually follow real world architecture built at a human scale. That means making sure rooms don't feel too small or too big for people. Proper scaling is very important when working in a realistic style, because it makes the level feel more plausible -- it will feel more like a real place built by humans and inhabited by humans.

But imagine a game about being a cat, it would need some rooms that feel too big (because cats are smaller than humans.) Or imagine a car game -- it would need houses that feel too small (because cars are bigger than humans).

For games we "build at the player's scale", which may or may not be a human scale.

Video game scale is weird

The human scale is helpful, but video game spaces are not human. Video games often rely on an exaggerated sense of scale that does not correspond to any consistent real world measure.

In Doom, the player ("Doomguy") is 32 pixels wide, which translates to 1 meter wide. The shotgun is 63 pixels long, which is almost 2 meters (6 ft. 5 in.) long. Oh, and Doomguy runs at 30 units per frame, which is 40-50 MPH depending on which hotly debated scaling system you use. Either way, Doom scale doesn't make any sense, but it feels nice to play anyway.

In The Elder Scrolls: Skyrim, the world area is about 14 square miles, the tallest mountain is 766.5 meters above sea level, and the player runs (not sprints) at 11.7 mph without tiring. That means the entire "province" of Skyrim is half the size of Manhattan, the tallest mountain is shorter than the Burj Khalifa in Dubai, and every peasant is a high performance marathon runner. By any real world measure, it is very small and weird. And yet, Skyrim somehow still feels like a large vast natural landscape with a huge alpine mountain, populated by typical people of average fitness.

So when building video game spaces, remember that it should merely "feel" realistic as part of the mood, atmosphere, or aesthetic.

But if you approach level design too much like real world architecture, then ironically, your levels will feel too big, complicated, and implausible.

Don't follow real world scale.

Scale advice

• "Big" levels feel big, but are actually much smaller than in the real world.

• Players regularly move at very fast speeds... that feel normal in the game world.

• The game world feels internally consistent, even if the specific numbers seem silly to think about.

Unfortunately it's not scientific. You'll only know "it feels right" when you're walking around inside your level and playtesting.

That's why we recommend considering metrics at the blockout phase. You can't really test metrics with a paper layout, but if you wait too late (like after an art pass) then it will be very time-consuming to make big changes. You might even have to redesign the entire level.

Basic metrics

In theory, you can build your game or level at whatever scale you want. In practice, physics implementations assume a default scale, and asset stores supply environment art at a common humanoid scale. Following convention is... convenient.

Common player sizes

• Bounding box / collision sizes are usually much bigger than the actual character model. This bigger thicker size is for more stable movement. It's not a hitbox.

• Note that in-game eye height is usually below where actual eyes would be. Many FPS games put your eyes in your neck or chest. For more detail, see Metrics of perception - Camera height.

• * Quake, Half-Life, Source Engine / "Quake lineage" games use "inches" lightly (eg. in Team Fortress 2, everyone is 7 feet tall) and don't really make sense.

Common modern architecture dimensions

• The minimum hallway width should be at least double the player width. Even then, it will feel a bit narrow and uncomfortable.

• Modern doors should be narrower than hallways to allow space for a door frame.

• Modern flights of stairs have landings / platforms every 12-16 steps. Long or tall stairs will feel industrial, monumental, or otherwise non-domestic.

• Modern stairs should follow a 30-35 degree slope. arctan(7/11) = 32 degrees

• In Source 1 and older engines, level grids followed power-of-two numbers. Unity users often prefer 1.0's and 10's. Unreal users are split based on their age. Use whatever number roundings that you and your collaborators prefer.

• When building in a realistic modern style, do some research beforehand: gather plans, blueprints, schematics, or even read through local building codes.

Deviating from conventional metrics

Deviate meaningfully from typical metrics, proportions, and spaces.

Steep stairs will feel more harrowing and awkward than shallow stairs, and if that's the experience goal you want, then go for it. Often the best part of a level is the uncommon and unexpected. But to make something feel unusual, you must contrast it with enough usual common elements first.

What if the level is stylized / isn't realistic?

Even with stylized non-realistic art styles, metrics still need to feel internally consistent to the game world / design norms. For example, the steps in Quake 1 maps are often 16 inches tall, but the rest of the game world is appropriately big and chunky to match that proportion.

What if the level isn't modern?

Do research on the type of architecture you're using. For example, traditional Japanese architecture uses the ken (間) proportional system. Italian Renaissance architects used the golden ratio (1 : 1.618) in the vein of Ancient Greek and Roman architecture, etc.

For more information on planning, see Pre-production and Research.

How to use metrics

Some ways to bring metrics into your level design process:

• Build a metrics zoo test map

  • Prototype core player metrics

  • Use world-aligned grid textures

  • Test modular kit

• Metrics debug tools, editor previews

• Codify shared building metrics

Build a metrics zoo test map

Build a "metrics zoo" / "metrics playground" developer-only test map that exhibits the different gameplay objects, modules, and prefabs across all your levels.

1. Blockout. Make a simple level with various rooms, hallways, and floor heights.

2. Playtest for usability. Walk around with gravity, jump on objects, be a general nuisance.

3. Playtest for feel. Pay special attention to the sizes of walls, doors, and floors. Does the height of the stair steps feel right? Do doors feel too narrow or too short?

4. Iterate. After each metrics revision, playtest and walk through again. Repeat.

Some example core player physics metrics

• Basic movement

  • Walk speed, run speed

  • Maximum step height, maximum angle slope incline

  • Minimum hallway width, minimum ceiling height

• Jump, fall, drop, height

  • Maximum jump height, maximum jump distance, running jump distance

  • Gravity strength, fall speed

  • Fall damage height threshold for 1% damage vs 100% damage lethal fall

Use world-aligned grid textures

When blocking out, use some kind of world texture with a little surface detail.

• Do not use flat solid colors, these lose visual scale and depth cues while prototyping.

• Do not use detailed textures, these are visually busy, can distort scale, and are distracting.

We recommend using placeholder grid / checkerboard textures, anything with repeating lines at regular intervals that will help you eyeball distances and proportions as you build your blockout. Check our Resources for links to various prototyping texture packs.

While building, make sure the texture scale stays constant and independent of the 3D object scale.

To ensure a constant scale in your game engine:

• Quake / Half-Life / Source: make sure your level editor's "texture lock" is disabled, and set all brush faces to use world axis alignment. For TrenchBroom, see "TrenchBroom Manual: Working With Textures".

• Any modern game engine: configure the shader or material to use a triplanar projection with a tilting grid texture. This has the added benefit of automating UV unwraps across all objects, so you can focus more on geometry rather than texture alignment.

  • Unity: visual shader editor has a built-in Triplanar Node, or to go deep on coding your own shader see Ben Golus' post "Normal Mapping for a Triplanar Shader"

  • Unreal: material editor has a built-in WorldAlignedTexture Node, or to go deep into coding your own shader see Ryan DowlingSoka's post "Triplanar, Dithered Triplanar, and Biplanar Mapping in Unreal"

Test modular kit metrics

With a modular kit, you build a level with pre-made tiles and components that snap together.

What's the best size for a doorway? You don't have to worry anymore; that metric is baked into the doorway module. In modular construction, you measure once, and basically copy-and-paste that measurement across the entire level.

However, it is common to iterate on a modular kit on the fly -- adjusting tile sizes and measurements, refining and art passing modules later, etc. If you make big changes to the modular kit's design assumptions, such as tile grid size or granularity, then you will likely have to rebuild everything because the new kit won't work for levels that used the old kit.

The best way to test the metrics and scale of a modular kit is to use it! Blockout with it, try to build hallways that loop back on themselves, then playtest and walk around in it.

For about modular construction, see Modular kit design.

Metrics debug tools and guides

For action games with a heavy focus on player movement and platforming, it can be difficult to predict whether a wall is too high to jump, or a gap too wide, etc. An in-editor metrics measurement tool / visualization / trajectory preview can help resolve these ambiguities.

For Psychonauts 2, developers implemented a "jump preview" editor visualization so that level designers could easily predict a player's jump arc:

"The jump metrics tool is designed to be placed into the level when the game is not running to visualize Raz’s jump arc. This can be used to help place objects along an action path so they’ll be reachable by the intended jump type. The tool runs the jump simulation code at edit time, while the game is not running. Any changes to the jump tuning would be reflected by the preview arc."

-- Devin Kelly-Sneed, "Behind the Code: Designing Raz's Jump"

Codify shared building metrics

Beyond player metrics like maximum movement speed or jump height, you can also set recommended building metrics for yourself as a guideline -- and these shared guidelines across a design team can help make construction and collaboration easier. Some examples:

Combat metrics

• What is "close range" or "long range"? How big is a "big fight"?

• How tall is cover, how high should a window be?

Pacing metrics

• How long should each map or chapter be altogether? How far should the exit be, how much backtracking should be allowed?

• How far apart should the various map objectives be?

• How many resources or rewards can be placed in each stretch of the level?

Storytelling metrics

• How long is this NPC dialogue, and how far can the player travel while they are talking?

• How many narrative assets such as audio logs, readables, codex collectibles, etc. should be placed in a given part of the level?

Optimization metrics

• If we are streaming in chunks of level data, how big should each chunk be?

• How long does the level transition hallway or sequence need to be, to mask the level streaming time?

• How many different meshes, materials, or shaders, can we load for each level?

Some example combat metrics for Team Fortress 2, based on observation and analysis from the TF2maps community:

• Close range weapons: 256 units or less

• Medium range weapons: 1024 or less (and projectiles become easier to dodge, etc)

• Rocket spam and snipers: 2048

• Maximum drop height without fall damage: 256

These metrics are reflected in the arena design of Team Fortress 2 maps like Badwater Basin below. Most drops are 256 units high, with maybe a few drops at 320 to apply a small penalty for any falling players. However there aren't many 512 unit drops because that would limit player options too harshly, and also render close range weapons much less effective.

For examples of detailed metrics, see Doom metrics and Quake metrics.

Metrics of perception

What affects the player's perception of size and speed? Basically anything on the screen.

Camera height

"Some animals are lightning fast. Others are pitifully slow. But slow and fast are relative terms. Four miles per hour doesn't feel very fast to a human: it's approximately one body-length per second. But to a small insect it's approximately 100 body-lengths per second. A human traveling that fast would be going 20,000 miles per hour! This is why some animals' brains process visual stimuli much faster than ours, and why they have better reflexes (think about how hard it is to swat a fly). What does it feel like to comprehend the world at such speeds?" -- Alex Reisner, creator of SpeedOfAnimals.com

If you are prototyping a game, finalize the camera height / eye offset as soon as possible, because it will affect how big the entire game world and all its characters feel. The camera height is one of the key variables that affect the player's sense of a virtual body, because "slow" and "fast" are relative to body size.

For suggested eye heights, see the Basic metrics section above.

Local reference points, composition, texturing, fog

TODO: more local context, objects along the line of movement

Camera field of view (FOV)

Your sense of perceived speed depends on how much you are seeing and the rate of visual change within your vision. A high field of view (FOV) will zoom-out to produce fisheye distortion that increases apparent speeds at the edge of the screen (periphery).

Note that there are three ways to measure field of view -- horizontal, vertical, and diagonal -- but for games we generally refer to the horizontal FOV so that it scales well with widescreen setups.

Some players prefer high FOVs because they can see more of the world around them and gain better situational awareness, and there have also been studies that suggest a high FOV helps some players mitigate motion sickness while playing first person games.

On top of the base FOV setting, many third person action games shift to a higher FOV when the player sprints or accelerates. This FOV change gives a greater sensation of speed, even if the actual speed increase is negligible. In contrast, some virtual reality games purposely occlude and mask peripheral vision to decrease this sense of acceleration and mitigate VR sickness.

The default Quake (1996) field of view was 90 degrees, and competitive players often set their camera to 120 degrees so that they could see more of their surroundings. This made more sense when CRT monitors and TVs had a 4:3 aspect ratio, but feels different on a modern 16:9 widescreen display.

Today, most commercial 3D action games tend to offer accessibility settings with an FOV range slider for users to adjust the camera to their personal comfort, usually between 60 and 90 degrees, but default to a narrower FOV. A high FOV distorts the image in an unrealistic / unflattering way. Meanwhile, a lower FOV zooms-in and "flattens" the view depth, giving a greater sense of closeness and intimacy with the game world. If your game is about looking up-close at characters' faces, you would err on a narrower FOV so that the faces don't look strange or distant.

Portrait photographers often try different lens focal lengths to get the most flattering distortion and closeness with their subject. 20 mm (84° FOV) narrows the face with extreme proximity, 70-100 mm (29-20° FOV) is more naturalistic, and a long 200 mm (10° FOV) telephoto lens widens and flattens the face with a longer distance from the subject.

Sound design, HUD / UI, and NPC AI as speed metrics

The brain processes all senses together. Vision isn't just a matter of seeing, it is also a matter of sound, memory, knowledge, etc.

To demonstrate, there is a well-known experiment in game feel involving two identical circles sliding past each other to swap places. Without any sound, the circles will appear to be just exchanging positions, but if you play a collision sound when the circles intersect, then the circles will appear to collide and bounce off each other instead.

For a more direct example of how other stimuli affect our sense of speed, consider the ZX Spectrum port of Super Hang-On (1987). In this single player motorcycle racing game, the "turbo" button temporarily increases the player's max speed limit to 324 km/h, beyond the normal limit of 280 km/h. However, the game does not render faster nor does the player actually move faster; instead, the engine sound gets louder, numbers get bigger, and opponent NPCs slow down. These contextual cues combine to give a faster feel without actually adjusting the camera nor the player.

The church of metrics: "rational game design"

A game industry practice known as rational game design (RGD) / rational level design (RLD) focuses on measuring metrics to plot parameterized game mechanics against game difficulty in a coordinate space, with the goal of mapping out a game's possibility space in a quantified way.

Basically, give everything in the game a number, and then add up all the numbers to help guide your design decisions. If you don't like how the numbers add up, change them. This general workflow looks like:

1. Design game mechanics, and identify various parameters and modifiers.

2. Plan levels across the entire game with different combinations of mechanics and modifiers.

3. Quantify parameters for each mechanic (e.g. 0% difficulty jump is a 3 m wide gap between platforms, while a 100% difficulty jump is 10 m wide, etc.)

4. Iterate on each level's metrics to conform to desired difficulty scores and target values.

Criticism of rational game design

Supporters claim this design method represents a more "rational" and "scientific" approach to plotting and pacing the player experience. Unfortunately there are very few public resources on how to practice RGD, which is convenient because that means we can't actually evaluate its methodology or claims -- its evangelists can always claim that we haven't read the proper holy scriptures, and so we don't truly understand how to practice it. What little documentation here is cobbled together from the few articles and resources available online, with much of the material guarded as a trade secret by Ubisoft.

We caution against any uncritical belief in any single design method, especially something that touts supposed objectivity / ahistorical understandings of "form follows function." And we remain skeptical of any method that claims to capture conceptual depth with a number. The more technical the design formula, the more formulaic the design will be -- a common critique leveled at many Ubisoft games.

But if RGD / RLD serves as a useful design tool that yields good results for the type of game you're making, then there's no harm in using it with moderation. Collecting data, measuring spaces and distances, and playtesting are all certainly good design practices. Just don't drink too much of this rationality Kool-Aid.

To review...

Metrics are useful measurements to help blockout levels at a consistent scale. If a level feels weird or awkward to play, it may need better attention to metrics.

To recreate a real world place, don't copy the exact measurements -- video game scale is often weird and inconsistent. Instead, scale depends more on your intuition of "what feels right."

To use metrics, blockout a metrics test map / zoo / playground with simple gridded textures, and then walk around in it. If you're using a modular kit, then build a test level and walk around in it, to make sure the modules are at the right scale and dimensions you need.

The most important level design metrics to consider are usually player physics metrics like movement speed, stair step height, and jump distance. You'll also probably want common building metrics like typical doorway width, window height, wall height, etc.

Metrics reflect human perception, which is affected by anything on screen or even game audio.

Many metrics are just "good numbers" and guidelines you give yourself, they're not a scientific infallible way to perfect level design. Metrics are useful, but metrics are not magic.

Now what?

• Use metrics to build a blockout.

Further reading on metrics

• Examples of detailed metrics breakdowns:

  • Doom metrics

  • Quake metrics

• TF2Maps.net: "Scale and your map" guide by Grazr. Even if you're not making a Team Fortress 2 map, it is useful to see how they derive their suggested metrics here.

• Source Engine / Half-Life 2 metrics (Valve Developer wiki)

When doing a blockout or art pass, you might use a modular kit -- a 3D tileset of modules, meshes designed to snap together. This comes from the real-life building practice of prefabricated modular construction.

If you're new to 3D modeling or level design, then don't attempt to design your own kit yet. Many considerations go into designing a robust kit. Your first kit will probably be very hard to use, especially if you haven't learned from using other peoples' kits already. Instead, see our list of Resources to download a free pre-made modular kit suitable for prototyping.

But if you have experience in level design and 3D tools, then you are ready. On this page we will detail the many considerations and best practices for designing your own modular kit.

1. Pre-production

do you have a lot of level designer time, and need to reuse limited art assets many times?

planning kit percentages, decide what is most important first

how granular / chunky should the kit be? depends on what level designers need / want, more customization OR faster prototyping time?

2. Blockout the kit

DO NOT MAKE VARIANTS YET, just build basic pieces first

Prototype metrics

decide on your units... build to a human scale

Common module types

• floor, ceiling, wall, doorway (single), doorway (double), window, corner

• platforms (landscape)

• glue

• flanges and arches (caves)

• shells (caves)

Define a footprint

stay in footprint / bounding box

Asset naming

pay attention to naming, level designers and kit artists need to agree on a pattern

Epic's recommended naming scheme?

3. Metrics: stress test the kit

do NOT just stop at an ideal case

collision is one of the biggest issues, make sure you can't get stuck?

Loopback Test

Stack Test

can you build multi-level environments? floors should NOT be paper thin, they have thickness and mass just like walls

Gap Test

do you have enough glue for interior kits with off-angle construction? essential for tunnel or cave kits

4. Art pass the kit, make variants

texture variations go a long way, and are "cheap" -- you don't have to test and validate the module geometry again

consider building a tool for easily swapping in module types, which will make art pass and remesh much faster

Sources

As mentioned at the top of this page, much of the info here comes from two excellent GDC talks by Joel Burgess, which are a must-read for anyone doing modular construction:

• "GDC 2013: Skyrim's Modular Approach to Level Design" (slides) is the "101" introductory talk to modular kit design considerations, and how they approached building out levels for Fallout 3 and Skyrim, while avoiding the worst of Oblivion's sins.

• "GDC 2016: The Modular Level Design of Fallout 4" (slides, video) is the "201" intermediate talk that elaborates on what they did for Skyrim's modular kits, and expanded when building Fallout 4.

Further reading

• "Modular Level and Component Design" by Lee Perry for Game Developer magazine, Nov 2002. A solid article by the lead level designer on Gears of War (2006), one of the first game dev writings to detail modular construction practices -- which only began emerging in the early 2000s when brush-based CSG construction began falling out of fashion. Included here mainly for historical interest; compare against History of the level designer.

Below are some core gameplay values and numbers that are useful for level design in Doom / Doom 2.

However, keep in mind this is an action game with aiming and dodging -- so the actual damage and damage per second (DPS) will depend heavily on enemy behavior, available cover, height changes, enemy composition, etc.

For more on what metrics are and why they're useful, see Metrics.

World metrics

Units and scale

Doom uses a grid with power-of-two numbers (e.g. 8, 16, 32, 64, 128, 256...) and textures are designed to work in increments of 8, 16, or 32.

Doom's graphics were designed for a 16:10 aspect ratio stretched vertically by 20% with non-square pixels for a 4:3 display. Today, this results in a lot of its art assets to appear vertically "squished". It also distorts any attempt at a coherent real world scale for Doom, which we can guess at: 16 horizontal Doom units = 10 vertical Doom units = 1 foot = 0.6 meters.

Maximum map size: to calculate distances, Doom uses 16-bit signed integers which have a maximum value of +/- 32767. However, if you actually built a map that stretched from -32767 to +32767 (across 65535 units!) and somehow tricked the engine into running it, then it would still break other distance calculations like a monster's line of sight, because a value of 65535 would overflow past +32767 to become 0. For best results, keep all map geometry within 16384 units of the (0, 0, 0) origin.

Weapon metrics

Doom randomly simulates damage values by rolling virtual dice with each hit. For the shotguns, there's an additional buckshot spread simulation where each pellet must connect with the hitbox for full damage.

The damage per second (DPS) is a rough estimate based on the fire rate multiplied by the average damage per shot.

Most maps begin with players killing low health enemies with the pistol and shotgun. Eventually the player relies more on the chaingun, super shotgun, and rocket launcher, while occasionally switching to the plasma gun for tougher enemies.

The rocket launcher, plasma gun, and BFG are usually less effective at very long range because of the lag built into their projectiles' travel time. At long distances, a monster can move out of the way before getting hit. Doom's autoaiming and randomized monster movement also means it can be tricky to lead shots. You can balance long range encounters toward the player's favor by placing monsters on pillars with no cover, limiting their ability to dodge the player's projectiles.

Monster metrics

Monster sight and hearing

Monsters can look in cardinal (N, E, S, W) and ordinal (NE, SE, SW, NW) directions, essentially in 45 degree increments. They have a 180 degree sight cone based on their initial facing, and can hear combat sounds based on areas bounded by linedefs set to block sound.

If set to "ambush" mode, monsters have a 360 degree sight cone and ignore sounds.

Monster movement

Minimum hallway size is given as (monster width + 2) x (monster height + 2) but in practice, your hallways should usually be much wider since monsters might "step" in larger increments, and monsters block other monsters. Narrow off-angle hallways will force monsters into slower zig-zag movements, because remember, they can only turn and move in 45 degree increments.

Stairs are tricky for monsters. In general, steps with long depths and shallow rises are always more dependable. Step height must always be 24 units or less (or else the monster won't cross) and minimum step depth / maximum slope is proportional to the monster width. For example, for a step that is 24u high, a trooper requires a step that is 33u deep (35 degree rise, 2:3 ratio) while a demon is wider so it requires 51u deep (25 degrees, 1:2 ratio). If you want to see the bounding box calculations yourself, see the PCheckPosition() and PTryMove() functions in p_map.c of the Doom source code.

To simplify building for monsters, we generally recommend:

• Minimum hallway size: 128 wide x 128 tall, mostly built orthogonally at 90 degree angles to align with the grid with occasional 45 degree angles.

• Stairway step size: 16 high x 64 deep (15 degree rise, 1:4 ratio) or 8 high x 32 deep.

Monster combat

Monsters will use melee attacks within 64 units of their target, though the Revenant will attempt to use a melee attack within 196 units even if the target is too far. If further than 64 units, then monsters with ranged attacks are more likely to use attack the closer they are to the player, up to a maximum distance of 2048 units. But the Arch-vile has a particularly dangerous ranged attack, so it will only attack within 896 units.

When hit, monsters have a random chance to be stunned in a pain state -- weapons with fast fire rates (chain gun, plasma gun) or multiple projectiles (shotgun, super shotgun) are particularly good at stunlocking monsters and interrupting their attacks.

For much more on monster behavior and debugging, see "Monster behavior" on The Doom Wiki.

Doom 1 monsters

Doom 2 monsters

Doom 2 includes all the Doom 1 monsters, and added more mid tier monsters designed to survive longer and interact with other monsters.

Doom monster design analysis

An influential Doomworld poster Linguica offers these helpful design thoughts on Doom monster design:

... Doom 2 monsters are great. They nearly all enhance the gameplay along one or more of three axes:

• time awareness - what is happening that I need to immediately address?

• immediate spatial awareness - what is in my close vicinity right now?

• general spatial awareness - what is the architectural layout, like walls, buildings, etc, in my area?

Doom 1 enemies were designed by what we would consider today to be novice FPS players, for whom basic movement and avoidance were challenging enough, without monsters making it much more difficult. However, as a more advanced FPS player, what tactical problems do you end up having while fighting them? Beyond getting boxed into a dead end, there is practically nothing preventing you from kiting Doom monsters indefinitely with little challenge. The only problems you would run into are being progressively plinked by shotgunners (troopers aren't even worth mentioning) or splash damage from a Cyberdemon's rockets (I'm discounting the Spiderdemon as a "normal" enemy here, but it is basically a giant rapid fire shotgunner in practice anyway, so whatever).

Now let's look at the new Doom 2 monsters and what they bring to the table:

• The Chaingunner is a "low-level" enemy that even a more advanced player needs to worry about, because especially en masse, they can really wreck you quickly. There is a time awareness factor because if you don't take care of them quickly, your health will be severely eroded. There is also an element of general spatial awareness in case you need to retreat behind cover.

• The Mancubus's offset fireballs mean you can't just rely on mindless strafing around projectiles and actually need to pay attention to them, or else you can end up running into one. This implicates your immediate spatial awareness. (Did you end up learning the "Mancubus dance" of strafing left-right-center? Then congratulations, the game taught you a new behavior.)

• The Arachnotron's rapid-fire plasma also implicates immediate spatial awareness because if you're doing any strafing back and forth, you have to watch out for "crossing the stream" and being hit in the process. Do you think it's a coincidence the Mancubus and Arachnotron are introduced in quick succession on the same level, which also includes no other enemies? I don't think so. Both enemies make you do more than the Doom 1 hallmark of simple strafing around a single fireball. The game is telling you, in case you haven't gotten the point yet, that this is not going to work anymore.

• The Pain Elemental is a annoying enemy, yes, but it is there purely to implicate your time awareness - it can't even hurt you directly! Its danger to you is directly proportional to how long you let it stick around.

• The Revenant's homing missiles, again, mean you can't just rely on mindless strafing around projectiles and actually need to pay attention to them, which involves your immediate spatial awareness, and you also need to keep mindful of available cover so you can get rid of the homing fireball, which involves general spatial awareness. You could even argue it involves your time awareness, since if you let a large number of homing fireballs accumulate, they could end up killing you instantly. (Or, perhaps, another enemy??)

• The Archvile's fire attack strongly implicates your general spatial awareness and time awareness - you need to know where cover is, and you need to get there immediately. Furthermore, its secondary behavior of resurrecting dead enemies involves yet more time awareness - if you leave it alone too long, it's going to being back all those enemies you already went to the effort to kill.

• The Hell Knight is the only new enemy that doesn't make the game tactically deeper along one of these axes. It's just a Baron of Hell without being quite so tedious to fight as a real Baron. This does give mappers a new option for when and how to use the Baron, though, which is a good thing.

-- Linguica on 28 August 2014 in thread "Doom 1 or Doom 2?" on Doomworld

Item Economy

Ammo pickups

Sources

• This page uses data from the Doom Wiki, under CC-BY-SA 4.0 International license.

• Map unit - The Doom Wiki

• Doom Metrics by Scott Ampoker

Below are some core gameplay values and numbers that are useful for level design in Quake 1.

Sometimes popular mods evolve new design norms in the community. After all, it's been more than 25 years since Quake's 1996 release. We note differences against default retail unmodded Quake ("Vanilla Quake").

For more on what metrics are and why they're useful, see Metrics.

For lots of Quake-specific info and links, see Quake resources.

Core player metrics

Quake uses the Quake Engine, obviously. So any basic scaling and metrics recommendations from the main Metrics page about Quake-lineage engines (Source Engine, etc) also apply here.

• 1 Quake unit (u) = 1 inch, but not really. There is no consistent scale.

• Z-axis up

• Quake uses a "power-of-two" grid. Most mappers blockout at a grid size of 64 or 32, then shift down to grid size 16 or 8 to start refining details. Use grid 4, 2, or 1 only for very small details.

• There's no crouching, sprinting, or +use key / interact button.

• *Falling damage is -5 health if the player hits the ground at any downward vertical speed faster than 650 units / second. In practice, this means fall damage rarely kills players. To kill the player from a tall drop, create a trigger_hurt or trigger_void (implemented in some mods).

  • Consider alternate ways of punishing the player for falling (monsters below, or water, slime, lava) or maybe even just let the player fall down, explore, and find their way back up again.

• ** Step height and slope assumes orthogonal (90 degree) axis-aligned surfaces. Quake collision uses Axis-Aligned Bounding Boxes (AABB) with no real support for rotations, so walking up a slope along a diagonal will feel much steeper.

  • Best practice is for ramps and slopes to be axis-aligned.

  • Many maps use Doom-style stepped terrain instead of smooth sloped terrain, partly to prevent physics problems. Monsters seem to like it better too.

• *** Jump distances and heights assume the absolute limits of a typical player.

  • Build for a shorter jump distance / height. Most players don't have perfect reflexes and timing.

  • Don't force players to ramp jump, grenade jump, or rocket jump on the critical path; reserve these specialty jumps for optional shortcuts or secrets.

  • Skilled players can use bunnyhopping to build-up more speed and distance, change direction in mid-air, etc. But don't worry about it. Let these players break your map, it's OK.

Common Quake building metrics

Player weapon metrics

Keep in mind this is an action game with aiming and dodging. The actual damage depends heavily on enemy behavior, available cover, height changes, enemy composition, etc. "Damage per second" is just a rough estimate.

• *Shotguns are a little complicated:

  • Damage is calculated via hitscan spread; for full damage, all hitscans must hit the target.

  • The effective range is much lower than 2048u.

  • Experienced Quake players use the default Shotgun more like a pistol since it has narrow spread, faster fire, and low ammo cost.

• ** Firing the Thunderbolt underwater causes an explosion that kills the player. This "discharge" consumes all Cells and inflicts cells x 35 damage. With 100 cells, that inflicts 3500 damage on anything that can see the player underwater... including the player.

  • but if the player has the invincibility powerup, they will survive.

Monster metrics

Monster collision

Quake has two collision systems. QBSP bakes collision hulls for basic movement clipping (entity vs world collisions), and then there's also hitboxes for moving objects (entity vs entity collisions). Both collision types use AABB (axis-aligned bounding box) shapes that cannot rotate.

There are three (3) world collision hull sizes:

• hull 0: 1x1x1 ("points" like projectiles, missiles)

  • but a missile's hitbox is 32x32x32 (via movetype FLYMISSILE / BOUNCEMISSILE)

  • and in-game, missiles don't even use this hull, and traverse the BSP instead

• hull 1: 32x32x56 ("small" players, humanoids)

• hull 2: 64x64x88 (large monsters)

Meanwhile, hitbox sizes vary slightly for every monster, and are listed in the table above.

ADVICE: HOW TO AVOID STUCK MONSTERS

Allow more extra space beyond the monster's size or else it will have trouble moving around: at least 16 units above ground / 16 units away from walls. Placing monsters slightly in the air is common.

Generally, place monsters in spaces at least double their width and height.

• For example, an Ogre (64 x 88) is better in a room that's 128+ wide and 176+ tall. For ease of construction, it's common to round up the room height to an increment of 64 (192, 256).

• Fiends and Spawns jump a lot and definitely need a 128+ tall room, ideally 256.

• But sometimes less monster movement is better, it all depends on the level.

Gibbing

Gibbing occurs when monsters get damaged a lot, all at once. If their health drops below a certain negative threshold quickly enough, then they turn into "gib" pieces instead of playing their death animation. This generally only happens when the player uses explosive weapons (grenade launcher, rocket launcher) or the quad damage powerup.

In vanilla Quake this mechanic only matters for zombies, who revive themselves unless they lose all their health (60) within a single frame and become gibbed. Again, the player will usually need explosives or a quad to kill the zombies permanently.

Infighting

Infighting happens when a monster damages a different monster type, which causes the hurt monster to target the other monster instead of the player.

For example, when an ogre's grenade accidentally hurts a fiend, the fiend will turn around and attack the ogre. But if an enforcer's laser hits another enforcer, no infighting will occur.

Melee-only monsters will never cause infighting. Only ranged monsters can.

Frequent infighting can make encounters feel too easy and unbalanced. But sometimes infighting makes for a more interesting battle. It all depends.

It is possible to design your map to require infighting by placing many monsters / less ammo. However, balancing these types of combat puzzles is complicated. Make sure to set a clear player expectation that infighting is required, otherwise players will be confused why there wasn't enough ammo. We recommend designing these situations only for Hard mode.

Skill level / difficulty levels

Quake has 4 difficulty levels: skill 0 "Easy", skill 1 "Normal", skill 2 "Hard", and skill 3 "Nightmare".

In general, health and damage do not scale automatically with difficulty level; skills 0-2 depend on mappers to manually adjust monster count and items.

The exception is skill 3 Nightmare, which enforces several code-level changes:

• 1996 vanilla Nightmare:

  • faster monster attack speed / frequency

  • less likely to stun monsters (enter "pain frame" animations)

  • Shambler lightning lasts 33% longer, for a maximum of 40 damage

  • Vore balls move 40% faster

• 2021 remaster / Copper Nightmare:

  • all 1996 Nightmare-specific changes reverted; Hard is now identical to Nightmare

  • except the player's maximum health is now 50 instead of 100

Item metrics

Ammo

There are three ways to get ammo: pickup an ammo item, pickup a weapon, or pickup a drop from a dead monster.

• All ammo item pickups can be flagged as "large", which doubles the ammo they provide.

• Gibbed monsters do not drop ammo.

• Army grunts die to 2 shotgun shots, but drop 5 shells, resulting in +3 "profit". If you use a lot of these monsters in the level, then balance your item_shell quantity accordingly.

• Ogres die to two rockets, but also drop two rockets, so using rockets on them is "free." Or alternatively, you can think of them as a way to exchange other ammo types for rockets.

  • For this reason, the rebalance mod Copper reduces the ogre drop to only 1 rocket.

  • But don't require players to use any of this "economic thinking." See Advice.

• Nails / spikes are the only ammo type that do not drop from monsters.

Armor

Player armor follows an overcomplicated formula, and honestly, we recommend just ignoring these details. It's not important. But for the sake of completeness:

• When the player gets hurt while wearing armor, they lose armor points (instead of health) based on the armor type's damage absorption rate.

  • example: Red Armor will absorb 80% of the damage you receive and subtract it from your armor points instead -- in practice, that means Red Armor rarely lasts very long, even though it seems like you get more of it.

• If you have too much of a better armor color, then you can't pick up a worse armor color.

  • example: if you have 114+ points of Red Armor, then you can't pickup Yellow Armor.

• Green is common in single player maps

• Yellow is considered to be pretty powerful, use carefully and sparingly

• Traditionally, Red is very rare, and reserved more as a reward for a secret

Powerups

• in vanilla Quake, the Biosuit and Ring are considered lackluster powerups with limited benefit that depend heavily on the level designer

• the Quad is probably the most interesting powerup

  • lets the player gib zombies with just the Axe or Shotgun

  • gibs enemies more often, which means they don't drop ammo

  • but also an opportunity to conserve ammo, e.g. use 1 rocket to clear an entire room

  • splash damage from grenades / rockets can easily backfire, killing the player instantly

Advice

• Quake monsters aren't very clever, which is part of the charm. Think of their movement more like semi-chaotic pinball rather than a tactical cover shooter.

• Give more ammo than the player will need. A first playthrough will waste ammo. If the player theoretically needs only 20 shotgun shells used with perfect economy and accuracy, then you should probably give at least 40.

For more general advice on designing combat, see Encounters.

Lastly, here's veteran Quake modder Matt "Lunaran" Breit with some Quake-specific advice and approaches for balancing encounters, item placement, and difficulty:

I think resisting "easier is just fewer monsters" design is the right way to go. There should be fewer monsters, just not many fewer, because too few is just plain boring. We're all Quake Experts after 20 years of this, so I think anyone playing custom maps on Easy in 2018 is doing so because they're doing it on a lunch break or a stolen evening away from the kids, and not because they can't handle more than one fiend at a time. Maybe we should think of it more as 'higher investment.' Besides, Quake gives you tons of unappreciated variables that you can tweak by skill, both obvious and subtle.

• Armor makes Quake significantly easier. More reds and yellows, more often, effectively extends the player's survivability in a given fight by hundreds of HP. More Greens, or stretches without armor at all, shrink the until-death buffer to little more than the player's current health. If you do want to keep the same monster loadout on all three skills, give the Easy player several Red and Yellow armors and the Hard player only one or two Greens. They'll feel like completely different games.

• Adding a vote for weapon pickups coming earlier or later. On Easy, the next big weapon might come before the next big encounter so the player can kick ass with it, on Medium it might be placed within it so the player has to engage to grab it, and on Hard it might only come as a reward after beating the fight entirely without it.

• Greater monster variety leads to more ways the player can be attacked at any one time, requiring juggling more variables to avoid damage and find the safe place to be standing at any given millisecond. A shambler and a vore together are harder to handle than a pair of either. Variety also raises the chances of infighting, however.

• The angles that enemies are presented from makes a difference. In front of the player is easier, flanking is harder, behind is bordering on unfair depending on circumstances. Below the player is a turkey shoot, eye level is straightforward, and monsters up high have a distinct advantage.

• Quantity of resources matters, of course. Bigger medkit pools clearly make the game easier, plentiful rockets can be splashed around while rare ones are only for emergencies, etc. Nail weapon DPS is higher than SG/SSG DPS and so on.

• Frequency of resources matters too. A steady drip lets the player feel secure, but isolated bursts create situations where the player has to stretch himself to get to the next 'island'. Depending on where he makes his errors, he might have to stretch pretty hard (eg those 'quicksave with 5 health left' or 'shambler axe dance or bust' moments). Feast-or-famine item placement can induce mild stockholm syndrome, leading to more positive reviews :)

• Unless you're using a lot of Enforcers, maybe provide all players an early Lightning Gun and simply vary the cells provided, as a way of dealing more or fewer 'get out of jail free' cards.

• Don't forget that the difficulty spawnflags are present on every entity. If you're using monster closets, vary the locations of the ambush triggers. Have the Hard ambushes happen when the player is in the worst possible position, and give them a leg up or more warning on easier skills (or even leave the closet open on Easy so there's no surprise at all). You might even duplicate the doors so you can set different 'speed' keys per skill, so the harder ambushes are an instant pants shitting and the easier ones are more like a countdown until the monsters come out, complete with early warning aggro sounds. Doors can be temporarily barred behind the player on hard skills while he is free to retreat from a fight on easier ones. Falling into a pit can be a mild backtracking inconvenience on easy skills but death by spikes on harder ones. How much room is there between nail shooters in this hallway? With careful use of triggerable lights and skill-specific trigger_relays, you can even use light and darkness against the player differently.

Getting crafty with what you change between difficulty levels can give you ideas for entire encounters, although don't rely on that too much for interest, because any given player is probably only going to experience one such permutation and thus won't realize the need to appreciate how different it is from any others.

Here's a method I've been using. It's really rough, and time consuming without a custom progs to do it for you, but it can be a helpful way to ground your estimates.

A box of: 25 or 50 nails = 225 or 450 damage 20 or 40 shells = 440/880 dmg 6 or 12 cells = 180/360 dmg

6/12 rockets are harder to judge because of 1) splash damage and 2) zombies, but let's say every rocket is 180dmg, for 1080/2160 dmg per box.

Total all the ammo you provide in the map (add 25 shells for the starter ammo, add 2 rockets per ogre and 5 shells per grunt and so on), and that's the max amount of hit point damage you are giving the player to deal. Total the starting health of all the monsters, and compare the two numbers.

Researching id maps and popular custom maps reveals an average 'custom' of about ~3:1 on Easy, ~2.2.:1 on Medium, and ~1.7:1 on Hard. the id maps are generally above that curve (4/3/2:1), and custom maps tend to fall below it(2.5/2.0/1.5:1).

Careful cheapskate shot-counters can finish a map with a ratio of about 1.3:1 ammo DP:monster HP, but most players will have to resort to the axe at some point and will complain of shortage. RPGSP1, which was greeted by universal reviews of "good but I ran out of ammo at the end" still had a ratio on skill 2 of 1.4:1 DP:HP.

There are lots of outliers to these curves, though, because so much of it comes down to how the level design enables the player to use the weapons, as well as exploit infighting, choke points, etc. Do rockets get spent one at a time on individual zombies or can they be used to gib crowds of knights for maximum ROI? or are they useless against herds of shamblers?

It also matters when the player gets the ammo. Ammo the player doesn't pick up or can't use is effectively not present in the map at all. Does it come too late to be used when it was really needed? does it come too early and get skipped? or partially wasted when picked up by a player who's already nearly maxed and getting too much at the wrong times? How much of that ammo is in secret areas?

Also keep in mind, you can toggle different platforms / walkways per difficulty, to make traversal easier or harder for the player. It's not just about item and monsters.

Further reading

• Copper - Changes is an epic 5000 word design essay on Quake's design and how modder Matt "Lunaran" Breit approached his influential rebalance mod Copper.

• We have a dedicated Quake resources page with a lot more info and links.

Sources

• QuakeWiki.org - see list of Monsters, list of Weapons, under GNU Free Documentation License v1.3+

• Quake Wiki (Fandom), under Creative Commons CC BY-SA 3.0 license

What is wayfinding?

Wayfinding is when a player finds where they are in the level (orientation) and/or finds a route toward their destination (navigation).

Most level designers aim for players to wayfind their way through the game world "naturally" in an "immersive" mood with minimal disorientation or frustration, while coinciding with the designer's intended critical path(s). This style balances naturalism and plausibility with subtlety. But more often, the goal is the performance of false subtlety -- helpful details that seem subtle (to someone, somewhere?) but are actually very obvious to most players.

(video embed?)

Wayfinding vs. guidance

Decades of industry level designer orthodoxy have claimed that the goal of level design is to manipulate / trick the player into "feeling smart" while "subconsciously" following the critical path... and to this, we say: no, stop! Can we quit thinking like this?

Deception is not a productive way to relate to players, and constitutes a false theory of mind for how players actually play:

Players often avoid the perceived critical path on purpose, to explore side areas and progress through the game at their own pace. In open world and multiplayer games, players crisscross the same space repeatedly and non-linearly, without one single static unchanging goal to always guide them towards. And then sometimes players just want to ignore gameplay and relax, hangout in a virtual space, spend time with friends, or even break the game and upload silly videos of goofs.

These are all valid ways to play. Let's move away from the weird (and creepy) prescriptive designer fantasy about secretly mind-controlling players into playing the way we want.

(image)

Instead, think of it this way: we are co-creating the game experience with players, and our goal is to provide tools and information to help them use our level.

Sometimes it's fun to stay on the critical path and sometimes it's fun to ignore it. Sometimes it's good to have a lot of information and sometimes it's better to have little information. Sometimes a narrow set of solutions can feel frustrating or artificial, and sometimes limiting options might result in a better experience.

We argue that player guidance is a flawed way to understand how people use levels and virtual spaces. "Guiding the player" fails to account for how players structure their own play. Meanwhile, wayfinding is standard terminology in architecture and usability design fields, and emphasizes the player's agency (way-FINDING) and active co-authoring of flow.

Wayfinding theory

There's been a lot of research across architecture, environmental psychology, behavioral geography, and design, into how people navigate spaces. While humans obviously traverse virtual spaces differently from real world spaces, we hope / imagine the mental process is similar.

Psychology

People form mental maps to conceptually represent a space.

• Paths - the roads used to move around

• Edges - roads which define the boundaries and breaks in continuity

• Districts - areas which share similar characteristics

• Nodes - strong intersection points of roads like squares or junctions

• Landmarks - easily identifiable entities which are used for point-referencing, usually physical objects

short term vs long term memory

Dead reckoning / path integration is a method for guessing and tracking where you are when no landmarks are available. You can estimate your current position by (1) orienting your starting position, and (2) extrapolating along your movement direction, speed, and travel time. Sailors and flight navigators heavily relied on dead reckoning calculations before radio and satellite, and even your phone regularly uses motion sensors and dead reckoning to estimate its location -- but even without math or technology, humans and animals regularly use a loose intuitive sense of dead reckoning.

Kevin Lynch's Image of the City

Strategies

• Track following: to rely on directional signs on the road

• Route following: to follow the rules given, such as a pre-planned route before the journey started

• Educated seeking: to use past experiences to draw logical conclusions on where to go

• Inference: to apply norms and expectations of where things are

• Screening: to systematically search the area for a helpful clue, though there may well not be any

• Aiming: to find a perceptible target and move in that specific direction

• Map reading: to use portable or stationary maps and help the user locate themselves

• Compassing: to navigate oneself with a figurative compass, such as the location of the sun or a landmark

• Social navigation: to follow the crowd and learn from other people’s actions

Signage theory

source https://fabrikbrands.com/the-wonders-of-wayfinding-design/

• Informational: These provide useful information on the place where the users are, such as free wifi, opening hours, etc.

• Directional: As the name indicates, these direct users with arrows saying which way to go for whichever purpose. These most often at junctions when the user must make a decision about the route.

• Identification: To help users recognize where they currently are, identification signs can be placed at the entrances of buildings, parks, etc. They symbolize the arrival to a destination.

• Regulatory: These let people know what they can and cannot do in a given area and are most frequently phrased negatively with the aim of creating a safe environment. Examples include “no smoking” or “restricted area”.

For example, when players exit the introductory train station in Half-Life 2, they witness a dramatic city view with deep composition that highlights The Citadel tower at the center of the city. To get the player to notice this significant landmark, Valve level designers pointed the door frame outwards into the city square, accompanied by a flock of birds suddenly flying upwards.

This example highlights three guidelines for scene composition:

1. Players don’t look upwards unless something draws their eye.

2. Players look in the direction they are moving.

3. Players focus on contrast (in color, shape, lighting, and movement.)

Wayfinding aids

Below is a table of various methods to convey navigation information to the player, starting from very subtle methods at the top, and descending to very obvious guidance methods toward the bottom. When possible, it's considered good practice to start at the top, and gradually incorporate less subtle methods in response to playtesting.

A wayfinding aid is anything that helps the player navigate where to go and how to understand the space.

• Positive reinforcement, attract the player to approach.

• Negative reinforcement, know where you can't go to understand where you can go.

We organized common wayfinding aids below by "% certainty" -- a non-scientific estimate of the proportion of players will likely perceive and use the aid.

Subtle (1-20% certainty)

Subtle wayfinding aids are best when overlaid as extra detail / polish with another more reliable wayfinding aid. Or alternatively, use solely these methods to make an area feel more unique or special for a small proportion of players (e.g. a very secret room). A vague nudge.

Coarse (35-60% certainty)

Coarse wayfinding methods rely on the player's prior knowledge or familiarity with common level design patterns. Common in "cinematic" or "immersive" games with an exploration fantasy. Use these methods to supplement a critical path, or to mark optional resources or secondary routes. A suggestion.

Situational (70-93% certainty)

Situational wayfinding methods will feel like the level designer strongly urging the player to follow a certain path or direction, but not forcing the player to do so. These methods include temporary camera control, situational UI pop-ups, direct threats and direct benefits for player progress. A heavy push.

Again, these methods cannot guarantee a specific behavior from the player.

For example, imagine a room with two exits, Exit #1 and Exit #2. There are many enemy NPCs in front of Exit #2. Will the player run toward these enemies to engage the enemies, and thus gravitate toward Exit #2 first? Or will the danger and threat of these enemies push the player away, towards Exit #1 instead? What if Exit #1 resembles a locked door, and the player has zero keys?

The intended "message" and resulting behavior will depend a lot on the player's current health and resources, prior encounters and patterns, overall level pacing, etc.

But here's what 93% of players will likely understand: those enemy NPCs were guarding Exit #2, and Exit #2 might be part of a critical path, and at some point the player may want to inspect Exit #2 more closely.

Direct (95-98% certainty)

As close as you can get to telling the player "you can't go here." Static, always-on, permanent game elements that the player can trust unambiguously as the voice of the game designer. A thick solid opaque unbreakable wall is a great way to urge the player "don't go through this wall", because they literally cannot... unless the player is a speedrunner, or if the player is wallhugging to try to find secrets, etc.

Advice for wayfinding

Note that some of these navigation aids are impossible to avoid, and some are even a bit beyond your control. That is because everything in the game is a wayfinding aid, and everything plays a role in guiding the player around the space. A wall makes a player to go around it; stairs promise another space above or below. Everything is information that the player takes in, all to varying degrees based on their mood, and none of it is foolproof. Like any other piece of information, these navigation aids can all be misunderstood, ignored, or overlooked.

We discuss some of these wayfinding aids in more detail below:

Ground composition

TODO: grab a slide from David Shaver's talk

Signage

CS:GO ironic bomb site signs... some graffiti is very subtle and not very effective probably

Breadcrumbing

highly dependent on flow?... maybe put this in the flow or encounter section imo

To review

Wayfinding is the practice of providing navigation aids for understanding the level layout. We urge level designers to reject notions of "guiding the player" -- instead, we argue that players play games in a variety of ways, and thus guide themselves. Strong wayfinding design help players construct mental maps to navigate and use a level more effectively.

Wayfinding signals a lot to players:

• Show the player "what they're supposed to do", let them follow an intended authorial critical path to progress

• Build trust with the player, promise the game world has been designed and remains readable with consistent navigation patterns

• Demonstrate conventional craftsmanship and production value

Against wayfinding

But what if the player doesn't want to follow "what they're supposed to do"? Or what if you don't want to build trust with the player, or you don't care about conventional craft and best practices?

Heavy use of landmarking, signposting, and breadcrumbing, will make a level feel like Disneyland. In many cases, this sense of controlled fantasy / "architecture of reassurance" can be desirable. The fake forest and fake wilderness of a theme park is common and appropriate for the typical commercial action-adventure themed game.

However, if a true sense of wilderness is desired, then the design may require a sort of anti-wayfinding. Ideally a natural and wild space shouldn't feel designed at all. If the idea of a totally unspoiled place is crucial to the project, then signposting and breadcrumbing will feel artificial and compromise the sense of place and authenticity.

Does your personal residence or home have landmarks and signage? Of course not -- and the privacy of personal navigation and memory is what makes that home belong to you. In contrast, Disneyland is unlivable.

Route-finding as mechanic

"[fr0g] clan official server 24/7 zk map (for stranger)" by Marek Kapolk is a first person anti-climbing game about carefully descending downwards. The level design often feels haphazard and unreadable, with heavily fragmented circulation and very little clear flow anywhere. The result is that the player feels like they are actively creating their own route and path, and less of an authored critical path. Bennett Foddy writes about its focus on route-finding and unique anti-wayfinding level design:

You do a lot of downward climbing in zk map, but the core activity is route-finding. You look across a haphazard pile of shapes, and your eyes trace possible ways down, imagining the lines of sight and obstructions along each meandering path. Route-finding is one part forward planning and two parts 'dead reckoning': making a choice and then figuring out how to get yourself one step out of the mess you just made. There is something magical about route-finding in a videogame, since videogame worlds are untrammeled, immaculate spaces. Every unguided turn you take is yours and yours alone.

The thing about making a game involving route-finding is you can't really get there by designing great routes. No matter how good your level design skills are, if the player is following a path you laid out for them, they aren't really route-finding at all. The player becomes too aware of your intentions, and their own autonomy becomes subsumed in them. As mkapolk explains:

My process for making the levels was to scatter geometry more or less randomly and then try to traverse it. Sometimes when I was going down a map if I thought that an area shouldn't be a dead end I'd add some more stuff to it, but that's about as far as it went.

You can construct a level that players can route-find through, but you can't design it... or to put it more precisely, you can't crack out the Good Game Design if you want players to experience route-finding. To pass through a well-designed level is a hike, not an expedition.

-- Bennett Foddy, "that's not fun: zk map for stranger"

Now what?

• Playtesting is crucial. Like a lot of psychology and design theory, most of this is wet bullshit until you verify and validate the design in a playtest.

Further reading

• Mapstalgia is a collection of maps and levels from video games, drawn from memory by players. The resulting drawings are often beautiful and wildly inaccurate. It's the closest thing we have to a player's mental map, and very interesting for imagining how players parse level design.

A playtest is when someone plays your level, and you watch whether your level works or not.

Playtesting is a foundational skill and process in game design. You should do it.

Why playtest?

Playtests are when you witness whether your level is "fun" or boring, clear or confusing, pleasant or annoying. If players hate your level, you should deal with that problem as soon as possible. "Playtest early and often."

Level design takes a long time, and it is natural to lose motivation during a project. Playtesting with someone else can boost your energy when you see someone else enjoying your terrible broken half-broken map, despite your misgivings.

The commercial game industry takes playtesting so seriously that they perform many phases of testing like user research, QA, test markets, usability, certification, etc. The biggest game companies even run their own dedicated playtesting labs.

Playtest formats

Each playtest has a different purpose, depending on the audience.

• Self-playtest: you test your own level.

  • You should playtest your level yourself every day, as part of the design process

  • Good for catching obvious problems (does the level load? is the doorway wide enough?)

• Critique: someone else (usually a fellow game dev) playtests your level and gives you knowledgeable feedback afterward. You can even have a conversation to brainstorm how to solve the problems.

  • Ideally, do this after every phase: after layout, blockout, scripting, and art pass, etc.

  • This is where a lot of learning and collaboration happens, very common in workplace

  • Good for getting help on "wicked problems" -- problems that are so complex that you can't even describe the problem, much less solve it

• Public playtest: someone else (ideally a "normal person" / member of the public, and not a developer) playtests the level.

  • Watching their behavior is more important than their feedback

  • Very important for catching hidden problems ("unknown unknowns")

  • But normies have a hard time playtesting blockouts, you'll likely have to art pass first

  • "Blind" playtest: a public playtest but with minimal guidance or prompting; pretend they are playing the level by themselves, alone, and see what they do

How to find playtesters

• Critiques: join an online level design community, ideally a forum or Discord with specific experience in the genre or game engine.

  • If possible, try to do an in-person playtest. See if there's a local International Game Developers Association (IGDA) chapter and/or indie meetup.

  • You need people with enough game design experience to contextualize their feedback appropriately -- people who know how to look past unfinished systems, obvious bugs, or placeholder art, and get to the real problems at the heart of the design.

• Public playtests: put your level in front of friends or family members, who are often socially / emotionally obligated to engage with your interests.

How to run a playtest

There are many ways to run a playtest session.

Below is a general process for an in-person public playtest:

1. Prepare for playtest

Before you playtest, you should figure out what you're testing! Treat it like a science experiment -- form a hypothesis and design a procedure.

Here's a basic playtest prep checklist:

• Scope: How much of the game / level are you testing? How long will it take?

  (example: "the first half of level 2, for 15 minutes max")

• Focus: Which parts are most important to test? What are your biggest questions?

  (e.g. "is the level exit clear enough? is the boss too difficult?")

• Briefing: What instructions will you give to the playtester? Any content warnings?

  (e.g. "use WASD to move, but don't press Space, I haven't fixed the jump yet... oh and the game has spiders, are you OK with spiders?")

• Data: What data are you collecting? How will you collect it?

  (e.g. use stopwatch app to time how long it takes for them to reach boss #1)

• Notes: Are you taking notes / recording observations? How, where?

  (e.g. write notes in notebook with a pen... oh no I forgot to bring a pen!...)

2. Brief playtester

Briefly introduce the concept and how long the playtest session will run, set some basic expectations to respect the playtester's time and energy. If the playtest session may contain jump scares, spiders or insects, flashing or flickering lights, sexual content, or graphic/explicit violence, then give a content warning and get the player's informed consent.

You can also tell the playtester what you want feedback on, and set some basic boundaries and focus for the playtest. A more specific briefing is very helpful if your playtester is a fellow designer, so they can tailor their feedback.

Example briefings:

• "use WASD and mouse... have fun"

• "I'm looking for combat ideas, tell me if the encounters are any good"

• "I haven't art passed the level yet so just focus on layout, these are all placeholders"

• "try to use the double jump a lot, I want to see if it works well for this level"

• "the layout is final and we can't change it for budget reasons, so please only give me feedback on enemy placement and pickups"

But keep it brief. Don't over-brief and don't dwell too much on disclaimers, especially if you're having other developers playtest. As designers we often feel insecure or anxious about how our work will be received, but We understand your levels will be unfinished or work-in-progress, with placeholder assets and bugs everywhere.

3. Observe playtest / collect telemetry

Most of the time should be spent watching the screen, taking notes about player behavior, reactions, or anything that needs fixing. Make sure to watch from a short distance behind the playtester(s), don’t just breathe over their shoulder.

Feel free to ask brief clarifying questions about the playtester's behavior or thought process, but resist the urge to interrupt the playtest to apologize or explain. Whenever the playtest goes wrong, let that discomfort and anxiety stew in your mind for a little while. However, if the playtester is definitely stuck or if the level is clearly just broken, then go ahead and intervene.

If it is not an in-person playtest, then ask the tester to record their screen and send a private link to the video upload. Or better yet, the playtester can broadcast a private live stream and take questions via chat or voice. For multiplayer maps, it is best to ask an existing player group, server, or clan to play on the map and allow observers.

Collecting data / telemetry

Track basic stats such as play time, deaths, win percentages, etc. Most of this stat collection does not require any special plugins or tools, just attentive note-taking.

Some games can automatically aggregate player data and visualize patterns, especially for multiplayer maps. Do players use the shotgun too often on a map intended for snipers, or are too many players dying outside the spawn room?

Telemetry is automatically collected data sent to a stats server that can track player behavior and display the data in a table / spreadsheet. Types of telemetry include:

• Data tables with combat stats for most common weapon, progression, or gear

• Heatmaps, top-down visualizations of where players frequently move, earn points, score kills, or die.

(TODO: heatmap example)

4. Debrief playtester

Ask for feedback and debrief the playtester.

Have a few survey questions ready for the playtester. You could ask comprehension questions ("Did you notice the orange light above the door? What do you think it meant?") or ask for more subjective opinions ("Did the boss fight feel easy or difficult?)

Be kind and respectful to your playtesters even if they're giving you feedback that you don't necessarily want to hear. If someone offers suggestions in good faith, smile and take it with grace no matter what. The suggestion itself might be a poor solution to a design problem, but the reason for the suggestion -- and the issue it highlights -- is almost always real and valid.

But also remember, the playtest session is not for the tester's benefit. If the playtester has a harmful attitude or gives abusive feedback, especially at an in-person playtest session, then you can totally just end the playtest and ask them to stop playing or talking.

5. Analyze playtest data

After the playtest, you need to figure out what the heck happened and what you're going to do about it.

"We do a lot of playtesting at our studio, and we have very good telemetry tools for recording data from the playtest that we can analyze after the fact.

This telemetry [pictured below] is from a pre-release 20 player playtest. [...] The bottom numbers show different armor sets on each row, but different days from left to right. These let me see patterns in terms of popularity of gear, where in the game it's popular, how long players stay on the same gear, when they switch... and I use these charts to identify clumping that might be unhealthy.

... At the start of the game there's not many armors available, so everybody is using one of the first three [...] But what's happening here? A lot of people are sticking on this Alfheim armor. Is that something I should look into? (Spoiler, we did nerf the Alfheim armor...) So it's about looking at the data to reinforce your expectations, to have a healthy spread."

-- Rob Meyer, "Reckoning With Fate: Combat Design at the Scale of Ragnarok" (2023) (YouTube)

How to be a useful playtester for someone else

The goal of a playtest is to help the designer predict whether their game works for their audience. It's not about the player's performance. Watch out for these common tendencies:

• New to games / game dev: too "nice", doesn't verbalize problems or frustrations

• Gamers: too nitpicky, exaggerates small problems, performs too much

Best practices for critiquing a fellow designer

Briefing

1. If you're unfamiliar with the genre or design style, say so.

2. Ask "what do you want feedback on?"

3. Ask "is there anything I should know before I start?"

Playing

• Think aloud, try to say what you're thinking constantly. No one can read your mind. At the same time, don't be offended if the dev ignores what you say.

• Don't nitpick unless the dev asks you to. If you break the game, tell the dev and then move on. Unless the dev asks you to repeat the game breaker for analysis, don't continue to break the game in the same way -- don't seek to embarrass or humiliate the dev.

• Be honest with expressing your emotions. If you like something, say so. If something confuses or surprises you, say so.

Debriefing

• Say what was memorable to you, positive or negative.

• Ask questions about intent and planned solutions, make it into a conversation.

Example playtesting processes

Playtesting at Valve

At Valve Software, level designers playtest every Friday across a variety of different playtester types (fellow designers, non-designers, non-dev staff, sometimes "open" public testing). Playtesting is fundamental to their design process and drives many of their decisions every week. Valve level designer Phil Co discusses their process:

On a typical project at Valve, the cabal (smaller subset of the team which consists of 6-8 people) works on a weekly cycle. Every Friday, we would playtest the section of the game that we're responsible for with someone from outside the team (even outside of the company depending on how far along we are). Everything we do during the week is focused on that Friday playtest. [emphasis added]

On Mondays, we have meetings where we decide what the goals are for the playtest and who is responsible for each task. [...] You might have to prototype a new mechanic which would involve a programmer, or you might have to perform an art pass of your level which would involve an artist.

We learn everything from playtests, however, so we might have too much of one particular gameplay and the playtester suffers from fatigue. We correct this issue in multiple ways.

For example, in [Half-Life 2] Episode 2, where the player gets the car on the bridge, we decided to add a combat high moment. Most of the level up to getting the car is combat but with just a few enemies at a time. It was made to be sort of a funhouse of zombies. We added the building where you pop the roof panels off so that Alyx can help you with the sniper rifle based on our playtesting feedback.

-- Phil Co, from "Building the levels: An exclusive interview with Valve level designer Phil Co" (Interlopers.net), circa 2013

Player data and analytics

https://twitter.com/MonsterclipRSPN/status/1268371601348685826

https://www.youtube.com/watch?v=Ij2nmt3EuJY

Further reading on playtesting

• GDC 2009: "Valve's Approach to Playtesting" (slides via gdcvault.com, free) by Mike Ambinder is the foundational industry text, covering playtesting fundamentals. Ambinder outlines the core bread and butter "direct observation" and iteration process at Valve, which is definitely the most helpful thing for the average level designer. There are two important historical notes: (1) data science / telemetry is now much more important in the industry, (2) biometrics like heartrate monitoring / brain wave recording / eye gaze tracking is rare in the industry, and other studios think it's "more trouble than it's worth" for level design. (Ambinder is an experimental psychologist, so of course he favors biometrics.)

• GDC 2017: "Playtesting - Avoiding Evil Data" (slide PDF, video via YouTube) by Adriaan de Jongh covers modern playtesting practices, with practical tips for how and when to run playtests. De Jongh argues "evil data" is "playtest results that are distracting, unclear, or misleading", so it's important to focus on frequent direct observation with a variety of playtester demographics. Don't bother with forms / surveys / questionnaires, instead send a template email to remote playtesters to record their screen and voice with Open Broadcaster Software (OBS) and then upload an unlisted video.

What's a player persona?

A player persona is a player's long term pattern of behavior and overall motivation for playing a game.

Classifying players under various personas / categories can help you study players and interpret playtest data.

In reality, players are never just one category. People are a complex combination of all these motivations and more.

Nonetheless, this reductive simplicity can be useful for making design decisions and imagining your audience. Ideally, every level can satisfy every persona.

(TODO: flesh this out a lot more)

Play style vs player persona

Huizinga's game of politics

• Player

• Cheater

• Spoil-Sport

Bartle's "gamer psychology"

The 1996 Bartle Test of Gamer Psychology was one of the earliest attempts to classify video game players. Bartle was studying player behavior in early text-based MMOs called MUDs, and sorted all these MUD players into four general categories:

• achievers want to score the most points, maximize game progress

• explorers want to discover new places or systems, map-out the game space

• killers focus on PvP / competitive conflict with other players, sometimes unfairly

• socializers focus on cooperation with other players, roleplaying and fan culture

(TODO: include 4 quadrant matrix image)