Skip to content

Vision: how The Crawl plays

Written 2026-10-27 from the project's own design docs.

The founder's picture of what playing The Crawl feels like: the camera, the controls, how crowded the world is, how players change it, and where the pressure comes from. Decided 2026-10-26.

How the vision got here

The founder's first one-line description put the game "somewhere between Fortnite and Hellgate: London", in a persistent, shared, simulated sandbox world. That was read as Fortnite-style building, a shrinking storm and dense 100-player fights. The founder said the reading was "way off" and wrote a fuller second version the same day. The second version is the current one; the first is kept only as a record.

Decided

Camera and controls

  • V1. Players see their character in first person or over-the-shoulder third person.
  • V2. Controls match modern first- and third-person shooters, RPGs and action RPGs, on keyboard and mouse or gamepad.

World scale

  • V3. The early floors are so large that meeting another player feels rare, but not unheard of. Floor 1 grows to fit its population; later floors are sized to the number of characters who reach them (see Worldspace).

Changing the world

  • V4. Players change the world through their character's abilities, not through building or demolition tools that everyone gets. A character can bring down a wall because they have a skill or item that can, not because the game gives everyone a demolish button.
  • V9. Destruction aims at the level of The Finals. The Finals is a multiplayer shooter by Embark Studios whose destructible buildings are the benchmark here. Its technology is Embark's own and can't be licensed, so this means matching the approach and quality, not copying it. The founder calls it a "pie-in-the-sky dream": it's the aim-high feature, and the one feature the founder would be willing to go without if everything else got built.

Loot

  • V10. Mostly random-stat loot, with a sparse set of hand-made items made as needed.

Time pressure

  • V5. The pressure is a ticking clock, not a shrinking circle. Each floor closes when its time runs out.
  • V6. A shrinking circle can still be used as a mechanic inside a floor while the clock ticks.

Combat

  • V7. A large variety of combat abilities: magic, melee, guns, skills, items and more.
  • V8. Abilities come from many sources: race, class, environment, equipment, deity, a promise from the dungeon's AI, and others still to be defined. Variety is a requirement, not a nice-to-have.
  • V11. Everything is numbers and modifiers, applied to whoever owns a thing and/or whatever it affects. That covers armour, a rocket, a banana peel, a freight train and a mad god's spell alike. This principle is developed on Characters and progression.

Readings (the assistant's analysis, not decisions)

  1. Sparse floors make the hardest technical problem easier. Sending every player's actions to every nearby player is the usual bottleneck for big online games. If each player has few others nearby, that cost stays low, so one floor can hold many players in total as long as they're spread out. The load tests later bore this out on one PC.
    • Hot spots are the exception. Stairwells, safe rooms, towns and bosses pull players together. They're where load needs testing and where limits or overflow handling will be needed.
  2. V4 means the world has to be changeable piece by piece. If any ability might break a wall, dig a hole, flood a room or start a fire, the world can't be static scenery. The server has to track walls, terrain and objects that can change.
    • Choosing how destruction works (breakable chunks, voxels, or scripted break points) is the biggest technical choice the vision forces.
    • The books suggest a rule: some things are indestructible by design (stairwells, for one), and each floor decides what can break.
    • What one player breaks stays broken for everyone on that floor for the season.
  3. V7 and V8 call for a data-driven ability system. With abilities coming from that many sources, they can't each be hand-coded. They should be assembled from shared parts (effects, targeting, costs, cooldowns, conditions), so a new race, deity or AI promise is mostly data. That choice is cheap early and very expensive to change later. V11 later confirmed this direction.
  4. V2 plus V8 creates a gamepad problem. A large, varied ability kit is easy on keyboard and mouse and hard on a gamepad. Ability wheels, loadout pages and modifier buttons are the usual answers; it needs designing.
  5. Aiming needs server-checked line traces. First-person and over-the-shoulder aiming both need the server to check each shot along a line, with lag compensation (rewinding positions to what the shooter saw).

Open (founder's call)

  1. Target population per region. "Rare but not unheard of" is a feel. Roughly how many players per region on Floor 1, and does that density hold as later floors shrink?
  2. Which things are indestructible on each floor. The destruction target is settled in principle; the per-floor list isn't.

Resolved along the way: the loot model (V10), destruction in principle (V9), and what persists after Floors 1–3 (see Worldspace).