If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.
Procedural stories in video games often induce a specific kind of delight. You’ll know when it hits — a realization that the code and algorithms of the game seem to be generating a coherent narrative from your own impulsive, seemingly chaotic actions. It’s what 2020’s viral sensation Blaseball and this year’s breakout indie hit, Wildermyth, share in common — two strikingly different games whose reactive stories are nevertheless cut from the very same cloth.
Players have grown accustomed to procedural generation in a spatial sense. Just look at the endless variations of levels that define games such as Hades in the ever-popular rogue-like genre and the infinite planets that populate the virtual universe of 2016’s No Man’s Sky. But procedural narratives are a different beast. (Distinct, it should be noted, from pre-written branching stories). They’re slippery, simulation-driven configurations of plot, setting, conflict, resolution, and people.
Drama, as video games continue to prove, is harder to convince players of than space itself, which makes procedural successes all the more eye-catching — from mainstream hits such as The Sims to cult classics like Rimworld. Now it feels like this sandbox approach to storytelling is starting to bear even greater narrative fruit.
If you ask game makers about the origins of procedural narratives, you’ll get nothing resembling a consensus. For some, it started with randomly generated dungeons of 1980’s Rogue; for others, choose-your-own-adventure books. Nate Austin, designer and programmer of tactical role-playing game Wildermyth, sees procedural storytelling as stemming from tabletop board games like Dungeon and Dragons — experiences that provide rules and a structure from which a vast number of narratives can spawn. In his view, and those of more than a few others, 2006’s Dwarf Fortress, a management game about dwarves seeking to colonize an austere, text-based world, is the inheritor to this particular genus of narrative design.
For newcomers, Dwarf Fortress can be intimidating. Beneath its mass of inscrutable ASCII icons lies a fiendishly complex simulation. On a basic level, its world is filled with flora, fauna, foes, and resources, plus, of course, your dwarves, all of whom have unique personalities. Your job is to ensure their happiness by building a colony that can satisfy their various needs and, thus, ensure the survival of the group. You might ace colonization itself, but then, suddenly, a gigantic monster kills half your group, which means you fail to bring in the harvest. Just like that, the colony is no more. “It sort of naturally creates these stories,” says Austin over Zoom. “And because you’re invested in the personalities all along, the drama happens in your head.”
With Dwarf Fortress and 2016’s RimWorld (which self-consciously builds on the former’s legacy), Austin suggests this is the closest video games have come to the freedom of tabletop experiences — “theater of the mind,” as he calls it. The designer has fond memories of such experiences, having played and dungeon-mastered Dungeons and Dragons sessions as a kid alongside his siblings Douglas Austin and Katie Austin (both are writers on the game). This narrative possibility is precisely what Austin sought to capture with Wildermyth, albeit in a more accessible style than either of those two games.
Wildermyth begins, as you might expect, with randomly generated characters. At first, they’re humble homesteaders but eventually grow into battle-hardened protectors of the realm. Events happen along the way: love, rivalries, children, and death. The most remarkable aspect of the game is that neither the precise order of these events nor their exact form is preordained. Wildermyth cooks up its story on the fly, and so while narratives follow a broad structure, they never quite land the same. Crucially, these characters are yours, so you care for them in a way that would perhaps surprise even the most curmudgeonly player.
The game’s magic, explains Austin, is in its “alternating layers of handcrafted and procedural content.” There’s the grand central narrative with a clearly defined beginning, middle, and end — procedural events that stem from the combat and personalities of your characters; then there’s the comic strips (dubbed “Library of Plays”) that bookend each event. Crucially, these “plays,” of which there are well over 100, are written in such a way that the code is able to seamlessly incorporate your heroes into them. The game succeeds both on a momentary basis — by turns charming, funny, suspenseful — and as a larger work of fiction, ebbing and flowing like a classic fantasy epic.
Wildermyth’s magic also stems from its approach to time. Campaigns typically last a century of in-game time, which means you not only see your characters age but the enemy advance ominously across the world map. Procedural RPG Unexplored 2 (currently in early access) works similarly. Every time your character dies, time and the game’s foes march forwards, and in this moment, you see the game’s whirring systems most clearly. “With procedural storytelling,” says Joris Dormans, director of Unexplored 2, “there’s the suggestion of a machine beneath the hood. You’re interacting directly with that, and in a way, you’re collaborating [with it] to create the story. I think that’s so powerful.”
Both Wildermyth and Unexplored 2 are products of what’s become a cottage industry of procedural storytellers across Europe and North America. Montreal’s Kitfox Games is both a developer of games like the wonderfully gloomy cult simulator, The Shrouded Isle, as well as a publisher. It’s set to release a visually updated version of Dwarf Fortress on Steam, part of an ongoing collaboration between co-founder Tanya Short and Dwarf Fortress designer Tarn Adams. (The two co-edited 2019’s Procedural Storytelling in Game Design — what amounts to a bible for aspiring writers and designers in the field.) Across the Atlantic, Emily Short (no relation to Kitfox’s Tanya) recently joined Failbetter Games as creative director, bringing a wealth of interactive fiction experience to the studio developing the romance-murder game Mask of the Rose.
Short’s CV includes both interactive fiction classics (2000’s Galatea asks you to have a conversation with a sculpture), as well as tools like Versu (axed in 2014 by its proprietor, Linden Labs, the studio behind Second Life). Like Dormans, whose work in video games stems from his PhD on emergent game design at the University of Amsterdam, Short has one foot in commercial game development and another in academia. She’s not alone — there’s a big crossover between the two worlds because studios are reluctant to fund costly research and development when there’s no guarantee of successful outcomes. Game makers often spend a few years cutting their teeth in the industry, moving into academia for research, before returning to commercial game development armed with a fresh set of narrative tools.
The major hub of this academic study is University of California, Santa Cruz, home to students and professors that cluster around its Expressive Intelligence Studio. Formed in 2006 by Michael Mateas, maker of the ambitious 2005 conversation simulator Façade, the work that emanates from the lab continually pushes the boundaries of AI and storytelling. Academics such as Max Kreminski are focusing their efforts on what they call emergent narrative, another name for the same kind of simulation-driven plots of Wildermyth and Dwarf Fortress. Over an audio call, they describe it as a “bottom-up” approach to narrative design — their job is to “find and bubble up” the interesting stories that fall out of the player’s interactions.
One way Kreminski is attempting this is through what they call “story sifting,” an approach that could give even greater shape, structure, and meaning to these procedural narratives. Think of it as the computer scanning (or sifting) in-game events to find interesting micro-stories — perhaps a lovers’ tryst or an escalating tale of revenge. These are surfaced to the player and then woven back into the game. The greatest challenge, explains Kreminski, is not in identifying these stories (he feeds the computer examples of what to look for) but matching events that have already happened with those that are in the process of emerging. If he can do this, these stories can be stitched together in such a way that they become a cohesive whole, a kind of plot-combo that stretches both in front of, and behind, the player.
Kreminski points to a few titles that have utilized something close to story sifting in the past. 2004’s The Sims 2 features “story trees,” which recognize sequences of events and nudge the player towards completing them. Social simulation Prom Week (which emerged from the Expressive Intelligence Studio in 2012) features characters that look at the history of its high-school world to influence next actions.
But perhaps the most in-depth implementation of story sifting is in 2020’s ongoing absurdist baseball simulator, Blaseball. Players bet on matches involving bizarre fictional teams (like the Baltimore Crabs) whose roster includes even weirder procedurally-generated characters (their traits can range from Shakespearian to anti-capitalist). The sifting happens in a few different ways, explains Cat Manning, narrative and design consultant on the game. At first, it was simply oral storytelling — players would have to see an event happen in real-time and then relay that to the community. But then, because the game spits out an eye-watering amount of data, players were having trouble keeping track — which is where the Society for Internet Blaseball Research stepped in. The fan community essentially developed a tool to let players watch replays of past games using data available on the site, what Manning calls a “backward story sifting.”
Blaseball now features its own “feed” on the site that surfaces important storylines to players, but a great deal of the story sifting still happens organically in the Discord chat. While most procedural stories occur within single-player games, Blaseball is a rare example of what happens when the scope is blown up to incorporate tens of thousands of players — almost like a mass hallucination. The deluge of fan art and even music is a testament to not just the stories its spreadsheet-like simulation generates but the way we as players are able to imagine color and detail into its world, just as readers do with literature.
You might be wondering why major studios, usually so quick to incorporate such innovations into their glossy blockbusters, haven’t yet jumped on procedural storytelling. The simple answer is that the variables involved become infinitely more complex and expensive at such scale. Still, there are a few exceptions: 2014’s action-adventure Middle-earth: Shadow of Mordor tracks your showdowns with the game’s Uruk mini-bosses through its Nemesis System. In turn, the game creates rivalry-focused mini-narratives inarguably more enjoyable than the main story itself. There’s also the State of Decay series, which features a cast of algorithmically generated zombie survivalists, the first a great deal more successfully than the second.
By far, the most ambitious blockbuster attempt at procedural storytelling is 2020’s Watch Dogs: Legion. Its hook, the “Play As Anyone” system, lets you recruit non-playable characters to a crack squad of hacker-activists seeking to liberate a grim, dystopian London. Every single non-playable character has a unique background — occupation, hobbies, relationships, felonies, special abilities — all of which are generated by the game’s “Census” system. The beauty of the game lies in assembling this squad according to your own specific preferences.
While the “Play As Anyone” system is easy to understand, its implementation was anything but, says Liz England, team lead game designer, over Zoom. She describes it as like picking up a rock only to discover there’s an entire civilization attached below. Because of its procedurality, whole aspects of production had to be relearned — animation, lip-synching, in-game lighting. “There’s already so many spinning plates shipping a game like Watch Dogs: Legion,” says England, “and then you say to everyone, ‘This tech you’ve been using, we’re gonna throw it out. This pipeline, we need to invent it from scratch.’ To make a game with a difficult concept to wrap your head around from a developer’s side and then to scale that to hundreds of people across multiple studios around the world — it’s very different to if you’re making an indie game.”
England, who recently joined a new studio headed up by State of Decay’s Jeff Strain, says there are other aspects of indie procedural storytelling that big-budget games can’t hope to match — at least not yet. Take medieval power simulator Crusader Kings 3, which generates text boxes with information specific to your playthrough. That becomes much harder when it’s something somebody actually has to say out loud. “The whole audio pipeline is just very expensive,” continues England. “It needs to be done early on because it has to be recorded with actors in other languages.” On a broader level, procedurality, especially within narratives, involves a relinquishing of control over the player’s experience. Making sure it’s one that can meet the expectations of famously exacting video game players — well, that’s tough.
As it stands, the indie world will continue to innovate, both in terms of games themselves and the tools used to create them. Emily Short, creative director of Failbetter Games, is excited about “socially and culturally democratizing” the space. “Having knowledge and tools be accessible, and having new people come in who want to experiment — this is part of the reason I write so much on my blog,” she says. “Even if specific innovations that I’m working on go nowhere, at least I’m equipping other people. It feels to me like there’s a huge space of unknown possibilities.”
Others echo Short’s sentiment. Because of the high level of technical expertise required to make these stories work, studios tend to have their own in-house software solutions. At the moment, says Tanya Short of Kitfox Games, it’s a case of “everyone building their own weird engine.” But she knows of one developer who’s about to start looking for private funding to develop their own tool (inspired, she says, by Versu). “There’s a need for it,” continues Short, “because there’s an emerging vocabulary set, but it’s basically industry jargon. Without words to describe [procedural storytelling], it’s very hard to discuss.” Alongside such tools, the Kitfox co-founder predicts machine-learning will buff up the audio and visual aspects of procedural storytelling, from voice-acting to art, so you’ll have “much higher production values on all indie games but especially those with procedurally-generated content.”
What’s clear is that procedural storytelling won’t wholesale replace straightforwardly human-authored plots anytime soon. For those worried about such a possibility, Short offers a thoughtful rebuttal. “That feels like it’s mistaking the pleasures of one thing for the pleasures of a different thing,” she says. Instead, players will continue to enjoy the sharply focused, linear narratives of titles such as those in the Uncharted franchise while being able to play games that convey the unmistakable, thrilling sensation of co-authoring a story with a machine.
While the likes of Wildermyth are part of a long procedural storytelling tradition, the field as a whole still feels as if it’s only on act one — the potential is as vast and varied as stories themselves.