Doom (1993)

Note: This page covers the original 1993 Doom by id Software. For the 2016 reboot, see doom-2016.

Summary

Doom (id Software, 1993) is a first-person shooter that established many conventions of the genre and of action game design broadly. Designed by John Romero (levels) and programmed by John Carmack (engine), with Sandy Petersen contributing a significant portion of the level set, Doom shipped via shareware — the first episode free, subsequent episodes paid. It became one of the most installed pieces of software in history within two years of release. Beyond its cultural impact, Doom is a design case study in emergent AI, environmental level craft, and distribution as player acquisition strategy (Barton, Vintage Games 2.0, see source-vintage-games-2).

Key ideas

• Emergent monster AI. Each enemy type in Doom operates on a small set of simple behaviour rules: wake on sight or sound, move toward target, attack on range, flee when hurt (Cyberdemon and Spider Mastermind only). These rules produce emergent encounters — particularly monster infighting, where enemies accidentally hit each other and begin attacking each other instead of the player. No designer scripted infighting; it emerged from the AI system. This makes every encounter subtly different and rewards player knowledge of enemy interactions.

• Height variation as level grammar. Romero’s levels are defined by height differentials that the engine (using BSP trees) could render efficiently. High ground provides tactical advantage; cramped corridors funnel enemies; open plazas allow rocket splash damage to become liability. Doom’s levels teach spatial reasoning through repetition — the player learns to read architectural language.

• Light and darkness as atmosphere and gameplay. The engine’s sector-based lighting meant that rooms could be pitch-dark or strobing. Darkness is both aesthetic and mechanical: players cannot see threats, and threats cannot always see players. Light variation is a design tool, not post-processing decoration.

• Environmental hazards as weapons. Lava floors, crushing ceilings, and barrels that explode when shot are not just obstacles — they are tools. Experienced players learn to herd enemies into hazards, multiplying their effectiveness without spending ammunition. This is an early expression of the principle that player creativity should be enabled by the environment, not just constrained by it.

• Shareware as onboarding design. Releasing Episode 1 (Knee-Deep in the Dead) as free shareware was simultaneously a distribution model and a designed onboarding experience. Episode 1 is the most polished, most carefully paced portion of the game. Players who finished it and wanted more had to purchase Episodes 2 and 3. The free episode was not a demo — it was the hook. This model influenced every subsequent id Software release and shaped the “free-to-play” logic that would emerge decades later.

• Modding as designed behaviour. id Software deliberately designed Doom to be moddable, releasing the WAD format as an accessible data container. The result was an enormous community of user-generated content that extended the game’s lifespan and created an entire secondary culture of level design. Design for moddability is an explicit structural choice.

In practice

Direct Unity/C# lessons from Doom’s design:

Emergent AI from simple rules. Avoid scripting specific encounter outcomes. Instead, give each enemy type 3–5 atomic behaviours (aggro, patrol, attack, flee, idle) and let them interact. Emergent results — enemies blocking each other, triggering each other, friendly fire — arise for free and make gameplay feel authored by the player rather than by the designer. See unity-gamemanager-pattern and state-machine pattern for implementation approaches.

Environmental affordances. Design levels with multiple “verbs” beyond movement: surfaces that damage, objects that explode, platforms that move. Players who discover these as tools feel clever. Players who use only basic movement feel competent. Both experiences are valid; the environment enables both.

Onboarding through free content. In episodic or demo structures, ensure the free content is your best content, not a stripped version. The first episode of Doom is demonstrably more refined than later episodes. This is counterintuitive commercially but correct experientially.

Evidence

Barton describes Doom’s monster AI as a defining example of emergent gameplay: “The monsters weren’t simply obstacles — they were actors in a drama, with behaviours that produced emergent encounters no designer had explicitly scripted” (Vintage Games 2.0, see source-vintage-games-2).

On shareware: Barton documents that Doom was installed on more computers than Windows 95 within its first two years — a statistic he attributes directly to the shareware model making friction-free trial possible.

On modding: Barton cites the WAD format as a deliberate architectural decision by Carmack, separating game data from game code so that level designers (including players) could work without touching the engine.

Implications

• Simple rules, complex outcomes is not just an AI principle — it applies to level geometry, item placement, and player tool sets. Design systems whose elements interact, rather than designing each encounter as a fixed setpiece.
• Distribution is design. How players first experience your game shapes their expectations, vocabulary, and tolerance for difficulty. Doom’s shareware model ensured the first impression was always the strongest content.
• Moddability extends lifespan. Games designed for modification create communities that maintain player bases well beyond the original release cycle. This has direct implications for educational or portfolio projects where community engagement matters.
• Height and light are primary, not secondary. In 3D environments, vertical variation and lighting change are among the most powerful affordances available. Treating them as decoration rather than gameplay elements wastes significant design potential.

Open questions

• Is monster infighting a feature or a bug that was ratified? If the latter, what does it suggest about playtesting-as-design-discovery?
• Romero and Petersen had distinct level design philosophies (Romero: intricate, maze-like; Petersen: open, alien). Does their stylistic difference within a single game strengthen or weaken the player experience?
• Doom’s WAD modding community predated the internet as a mass medium — it spread primarily on BBSs and shareware discs. How does distribution medium shape community formation around modding?

Related

doom-2016 · source-vintage-games-2 · bushnells-law · systemic-depth-elegance · state-machine pattern · game-loops · level-design · procedural-generation