The Gods Play Dice: Procedural Content in Blaseball
by Cat Manning and Sam Kabo Ashwell
Game designers have a saying about procedural generation: procgen gives you twice as much content for only twice the work. Some versions have it as twice the content for four times the work. Designers often get very excited about the potential of randomness, before realizing that it has serious costs and consequences down the line: very often, there’s an easier option. Heavy reliance on randomness and procedural generation is never something you have to do, in any corner of game design — you can always go with strict determinism, hand-crafted content, or just letting the player choose.
Blaseball uses a lot of randomness. It uses randomness to a greater extent than you would expect it to, even. It leaves some things up to chance which most designers would be really tempted to make deterministic, or scripted. What is this in service of? What does it accomplish?
A Mundane Wildness
Most obviously, randomness lets blaseball handle the huge volume of games and their details. There are 1,188 games in a season, and some quick back of the napkin math says there can’t be any fewer than 64,152 at-bats. (The actual number is always many times that.) With several hundred thousand generated events, most will be pretty routine. Most at-bats don’t make it on base; we tend to pay more attention to the ones that do, and remember them better. But because of the accumulated complexity of blaseball’s systems, the sim occasionally produces more exciting situations — sometimes by intentional design, sometimes from the combination of particular low-probability cases. And these, in turn, tend to have an outsized effect on the narrative we extract from the simulation.
There are multiple modes of randomness in blaseball with different effects on theme, strategy, the texture and meaning of the experience. So much of the game is driven by procedure and probability, a dice roll finding a particular value within a weighted distribution. Some of that randomness is the normal background noise of sports simulation: whether a batter strikes out or hits a single, if an at-bat turns into a sacrifice play. It forms the backbone of play, and all those moments add up to the final scoreline; a lot of them don’t end up mattering much, but any of them could. Without the pleasant background hum of games being played, events like incinerations or consumer attacks don’t hit as hard. To be exceptional or weird, events need normality to be contrasted against. Drama requires variations in intensity: peaks of intense excitement and surprise, plateaus of routine.
There’s an XKCD comic that points out that all sports commentary is just using the results of a random number generator to tell stories. Blaseball’s response to this is: yeah! Isn’t that great?
Every time a few low-probability things happen in a short period — two incinerations in two games, say — we all freak out because it feels as though Something Special is Happening. Randomness ensures that this kind of coincidence will show up now and then, but — because there’s so much mystery and uncertainty alongside the randomness — it’s never simple to separate signal from noise, and so we get to dream up lots of possibilities about what’s really going on.
Ambush Aesthetics
Procedural generation is excellent at fleshing out an aesthetic, sketching in details that add up to a possibility space. This is very good for Blaseball, which is suffused with weirdness: surreal double-entendres that are sometimes sparsely ominous, sometimes goofy, and often split the difference. Over the seasons, blaseball has added in more systems that add telling details and strange incidents over the course of a game.
Two that I want to talk about are coffee styles and grind tricks. There are three kinds of Coffee weather, which first showed up in the Coffee Cup, a one-off just-for-fun tournament. Players get “beaned” by certain styles of coffee — this has in-game effects, but there’s also a purely aesthetic, procedurally-generated description of the blend’s flavour notes. These descriptions feel interesting (and occasionally disgusting) even several seasons later. Players might get beaned with “a light roast with hints of lavender” or “a dark roast with candy-like sweetness”, or the more alarming “a cinnamon roast with mango and extra cheese” or “a blonde roast blending plum tomatoes with umami”. (Coffee 2 combinations, in particular, can produce delicious or intriguing blends like “milk chocolate with hazelnut” or “citrus with oak”, but many of its juxtapositions are memorable more for their wrongness.)
Grind rails first appeared as a stadium renovation in Season 15, and were immediately popular: half the league built one. Fans knew that it would let baserunners occasionally skip second base: what no one expected was the procedurally generated corpus of skate trick names. We had “normal” tricks like the Kickflip, the 1080, and the Handstand, but also a lot of blaseball flavor and inside jokes like the Bloodflip, the Umpdog, the Oh No, and the Blue Check. This works particularly well because we already have the sense of skate trick names as being a dense body of goofy in-jokes created by a specialist subculture’s terms of art — you know, just like Blaseball.
Since coffee and skate tricks each draw on a good-sized but finite corpus and simple generative rules, it’s possible to come across something you’ve seen before, or that feels close enough to it. This is a well-known feature of procgen: it offers small strange delights, but as one gets more familiar with its output, it sooner or later suffers from what Kate Compton calls the Oatmeal Problem. Slight variations on a theme might not offer perceptual uniqueness, which is what maintains surprise: “I can easily generate 10,000 bowls of plain oatmeal, with each oat being in a different position and different orientation, and mathematically speaking they will all be completely unique. But the user will likely just see a lot of oatmeal.” This is a risk here, because grind trick combinations aren’t hugely fruitful in combination; they’re just funny names. Even if they were real-world skate tricks, only specialists would have any idea whether, say, a 5–0 to indy nosebone was particularly impressive or interesting; and of course most of these are imaginary tricks that are harder to visualize. (Coffee flavors are less interesting on their own, but more conceptually productive: one can imagine what “notes of blood and cinnamon” might be like.)
Blaseball often avoids Oatmeal pitfalls because its most attention-grabbing elements are embedded within other systems of randomness; rail tricks are cool but they’re a sideshow. We have an expectation of routine, and the flavorful elements get to stand out against that. It is oatmeal, but there’s also raisins. And nuts. And pop-rocks. And edibles. And fish-hooks. Unpredictability factors in, too — we don’t know exactly when birds will attack, or when skate tricks or the Secret Base will trigger, so it keeps us in suspense. And if we lose that sense — routine is a sense of false security, and makes it more alarming when that comfortable blandness gets shattered.
Names are another place where blaseball gets to employ procedural generation to sketch out possibility. A lot can be said about this that we don’t have the space to cover here, but keep an eye out for Sam writing on this at some point. In the early days of blaseball, its pre-stocked names felt almost familiar: names we’d recognize on a jumbotron, with a proportion of Latine names suggestive of real-world baseball. But there are also stranger names like Emmett Internet and Blood Hamburger and Jessica Telephone. Names like those hit hardest and are most amusing when those players have teammates like Nagomi Nava and Esme Ramsey and Hiroto Wilcox and Arturo Huerta. If everyone’s a unique snowflake, the overall effect is just a blizzard: white-out.
Later seasons have introduced the Egg Pool, to which players at the Patreon’s $10 tier can submit their own names. This has shaped the sim’s output, giving us players like Gerund Pantheocide, Anathema Elemefayo, and Scratch Deleuze. [Editor’s Note: Cat is responsible for the last, and does not regret it.] Some of these submissions are “normal” names, probably to counterbalance the unwise joys of committing a particularly egregious Name Crime. But on the whole, they’ve tilted the corpus towards the Internets and Hamburgers and Telephones. This is not necessarily a bad thing — although it does indicate that there are a lot of competing aesthetics, reflecting contrasting visions for what blaseball is.
Truth Emerging From Her Well To Shame The Pies
Different visions of blaseball means that because the sim offers us no motives for the particulars of the stories it spits out, we get to identify and make our own; the joy of realizing when something’s especially, strikingly weird and pointing that out. These stories often become more compelling the more low-probability systems are involved, because the cumulative randomness required for each part of the chain to trigger is so relatively unlikely. As blaseball accumulates systems, there are more and more unexpected ways for them to interact.
There isn’t room to talk as much about emergence as Cat would like to, but here’s a brief example of how it produces evocative little fragments of story. Take the time Sparks Beans and Sandford Garner Charmed one another. On paper, this shouldn’t have been able to happen: only the Lovers possess the modification that lets them Charm other players into striking out or giving up walks. But since the Garages built Psychoacoustics, which can Echo other teams’ modifications for the duration of a game, the Garages were briefly Charming; after Sanford Garner charmed a string of Garages batters, Sparks returned the compliment. This is a cascade of actions of varying probabilities: that the teams would play under Feedback and in the Big Garage, that Charm would fire on a Garages player, that it would pick out the same players twice. What would have been routine turned remarkable because of several mildly-low-probability outcomes chaining together; Charm’s already an odd, one-off kind of system. Human brains enjoy identifying and interpreting this kind of interesting coincidence — and building systems that interact with one another makes for a lot more ways that can happen.
The Future is Fair Play
One of the things that audiences really care about in sports is fairness — or some version of it, anyway. For many, the distinctive quality of sports is a real contest — real doubt about the outcome, a real sense of accomplishment at success. (You also, if you’re gambling, want assurance that the game isn’t rigged.) All sports are, inherently, artificial dramas — there’s no prima facie reason why we should care if the man runs all the way around the square, or where the ball ends up, and in order for the action to make sense we have to buy in to the magic circle of the rules. But within that fictionalized space, we want the struggle to be real. That doesn’t mean we can’t enjoy pro wrestling kayfabe, or scripted sports dramas — but those are qualitatively different kinds of entertainment.
In most videogames, we can replay after death or failure. The narrative we ultimately get, this way, is accretive, built up in overlapping layers; many different attempts and alternate histories, overlaid, redundant, a little contradictory in the details. But too much of this can erode the sense of a single story, of the characters’ status as individuals. There are people who refuse to replay games that they were particularly moved by, because they feel as though it would betray the characters as they knew them. Heavy randomness plus heavy save-scumming can emphasize the artificiality of a game, and lower the sense of stakes: what does it matter if your ranger captain flubs a shot and gets wrecked by a chrysalid, if you can just keep going back to a one-turn autosave until she lives? This is why traditional roguelikes — which rely hugely on randomness — are often single-save. The consequence is that they can be extremely unforgiving; you can invest a whole lot in a character, only to have them die because of a moment’s carelessness or plain bad luck. Sound familiar?
Blaseball relies on the lasting consequences of randomness to impart narrative weight to the collective story unfolding. We can’t rewind the plot to get the result we want; we can try and undo our mistakes, but that carries an opportunity cost. We can only go forward and hope that we can find opportunities for redemption (or satisfyingly worse mistakes). Even though it’s unforgiving, it’s not punitive: because the consequences are left up to the roll of the dice, it’s never The Game Band coming for your faves.
Arbitrage
The English words arbiter and arbitration suggest a fair and trustworthy judge, someone who can be relied upon to decide a case. But we often dislike or distrust our judges and umpires, and thus we get the word arbitrary, as in capricious, tyrannical or random. When we make decisions by flipping a coin, we often do so by realising how much we hate the choice the coin made.
Blaseball is a story that involves character deaths and other unpleasant fates. In other media, particularly episodic forms where the audience has time to get attached to characters before the story’s over, character death tends to attract heated discussion and blame of creators. The thing is, fans often aren’t wrong about this, because from past experience we know that there are always lots of ulterior motives for this kind of major change. The actor and the showrunners didn’t get along, the writer wasn’t comfortable developing a queer character, the studio decided that this character wasn’t popular enough with the right demographic. Audiences — media-literate audiences, at least — have been trained to second-guess this kind of thing. There are long-established tropes here, and a tendency of fan-criticism to talk about deaths as if the authors killed the character. (The authors didn’t kill anyone! They wrote a story about a fictional character who fictionally died. But because the magic circle is permeable in one direction, it can get real leaky in the other.)
A consistent thing about blaseball storytelling is that TGB don’t make very many direct choices about individual characters. They mostly let the sim deal with the particulars — or, sometimes, the confused, conflicted choices of the fans. There are systems set up, slots open; but there’s always the possibility of being surprised at who ends up in the spotlight, or on the chopping-block.
I’m always perplexed when I hear people claim that the Season 10 JRPG-esque boss fight against the Peanut must have been staged, because we won, and the Game Band wouldn’t have let us lose. And I always think, are we playing the same game? We absolutely could have lost. All of the modifications the Hall Stars were laden with almost weren’t enough to push us over the top. Without that possibility — the world in which the team we’re rooting for loses — the game has no real threat, no real triumph. The strongest messages of a game are usually the ones expressed by its systems: a blaseball movie or book could not express these themes as strongly.
Blaseball is a game about inhumane systems, and their indifference to humanity; there is no prejudice or favoritism, and that means that everyone is vulnerable to the same horrors. On one level that makes it more palatable — you have some assurance that if a character you cared about dies, it wasn’t because of some kind of poorly-considered authorial fiat, prejudice-riddled focus-group intervention, or punishment for inciting discourse. An oft-cited principle of fandom criticism is that you shouldn’t give characters pointless, meaningless deaths; blaseball cuts directly against that. Very few deaths in blaseball have any obvious imparted Meaning; we have to figure that part out for ourselves.
Chaos Engines
Much of the meaning we draw from blaseball comes not from each individual event but from familiarity and context: when Annie Roland was incinerated — completely at random — it felt different because she was already notable for having survived Instability twice. Even though we only encounter blaseball’s system through its outputs, the stories in aggregate disclose the underlying structure and theme — and are often very on-the-nose about it. Blaseball’s themes have always been obviously political: it’s a story about the heartlessness of broken systems, about lack of security, about not being in control — and about what little power you’ve got being very difficult to wield. And the end-of-season elections — viewed with dread as often as excitement — are where that’s most apparent.
Randomness has always been built deep into blaseball elections, which is a source of hilarity and frustration. Any given system of democracy has to satisfy a lot of competing desiderata of ‘fairness’ — to make popular choices, to avoid widely-unpopular choices, to favour majorities while still representing minority interests, for every vote to matter equally, to let people vote for their real interests rather than force them to vote tactically, and so on. The raffle system of Blessings and Wills looks ridiculous and chaotic, but it performs pretty well on some of these standards. Anybody’s vote could be the one drawn out of the hat, so you can’t say your vote doesn’t matter, even if a large majority are voting for an option you hate. Of course, this system has a… mediocre-to-awful record at a lot of other things we want out of democracy. No voting system is perfect, of course, but it’s hugely silly to focus so strongly on this one aspect. Except that blaseball is a game, and voting is the most immediate way the player has of interacting with it: in that light, it makes a certain kind of sense to go with a system that values motivating participation above everything else.
Because it’s about incentivizing engagement, it often fails at satisfying other qualities of fairness. Quite often it fails spectacularly at almost all of them. And that’s kind of the point: blaseball isn’t meant to be fair, or is meant to be fair only in weird, dark-mirror senses. Blaseball democracy is an illustration of how democracy is only as good as its structure, its implementation and its underlying motives.
There are many kinds of horror, and — for all its extra elbows, blood-drinking and shark jump-scares — blaseball’s register of horror is a lot more Kafka than John Carpenter. It’s the horror of dealing with an inescapable system that you don’t fully understand, that resists understanding, that is not designed for your benefit and is deeply indifferent to your humanity. Randomness is everywhere, even where it doesn’t strictly need to be, and the result is that the system as a whole is partly obscured to everyone — sometimes even its designers. The story has shape, but the end is not yet written, the details are in flux, and some of the narrative authority is not granted to humans.
