Jelle’s Marble Runs, Blaseball, and the Joy of Chaos

“Creativity is the ability to introduce order into the randomness of nature.”

– Attributed to Eric Hoffer

This week marks the return of the Marble League, an Olympics-style competition hosted by marble racing YouTube channel Jelle’s Marble Runs. In it, a couple dozen teams of differently colored marbles, with their own team names and audience chants, participate in various sporting events like hurdles, collision derby, and underwater racing.

Simultaneously, the idle sports game Blaseball enters its 21st season, with teams including the Baltimore Crabs and Chicago Firefighters jockeying for championship glory through murderous rogue umpires, rains of peanuts, and the mysterious “ownership” who seems determined to feed the entire league to extradimensional sharks.

Niche but deeply passionate audiences have been built around both of these media franchises. People will pick a team, buy merchandise, and debate their relative merits and demerits in the social media of your choice. Team mythologies spring up constantly in both – the dominance and unique chant of the O’Rangers*, the perpetual mediocrity and tiny fanbase of the Miami Dale**, and so on for every team or name that appeared even briefly.

Passionate fanbases arising for odd media is nothing new. What interests me here is that Jelle’s Marble Runs, and Blaseball less so, are essentially random – not only does the audience have no control over what happens, neither do the creators, who set the mechanical (in the case of Jelle’s) or virtual (Blaseball) gears in motion like some kind of distant god and proceed to let everything run itself.

What makes this total chaos so fascinating, and how can we capture this in our game designs? Read on for some of my theories about it.

* I am an O’Rangers fan.

** I am also a Miami Dale fan.

Real Fake Athletes

There’s something to be said about narratives constructed through chance. This was a subject frequently attended upon by artists belonging to Dada and other surrealist art groups in the early 20th century, who would do activities like Exquisite Corpse that eliminated the ability for the people participating in it to meaningfully control what was being written. While surrealist movements have been, depending on your stance, either out of favor for a century or democratized into Internet memes, procedurally generated narratives remain extremely popular because of their authenticity.

What do I mean by authenticity? Let’s look at an example of a very popular procedural narrative: Sports. While some people watch sports regularly so they can vicariously succeed through their regional team, and others like to watch athletic accomplishments writ large, the overarching appeal of sports viewing is that whatever happens on the field isn’t prewritten.

A champion so dominant his batting average enters the realm of the godsan underdog team succeeding against all odds to prevent them from foldingan athlete overcoming hardship after hardship, only to shatter an existing world record. These things are the stuff of cliché in sports movies, and we go into it largely expecting these things to happen because that would be the most exciting option the writer could give us. We roll our eyes at them, in fact, because we feel the creator is pandering too much to our expectations. But these exact same things happening as the result of chaos – of people doing things at random with no overarching “plot” – they’re rare and therefore more meaningful and valuable.

Sports simulators like Jelle’s and Blaseball build the same environment of procedurally generated stories like “real person” sports do, but in a more condensed way. Notably, games are much shorter than their real-life equivalents would last: Jelle’s Marble Runs‘ equivalents to F1 races or Olympic events that might take most of a day only last five minutes, while the longest game of Blaseball ever played lasted around an hour and a half. For most nerds, who don’t have the patience or cultural buy-in for traditional sports, having these narratives become quicker to achieve and more likely to happen by virtue of sheer numbers is immensely appealing.

These games also possess humor elements that help liven up otherwise by-the-books gameplay. Jelle’s has announcer Greg Woods, who uses placid commentary that treats marbles on a track as flesh-and-blood athletes with mental states and training regimes. Blaseball goes much more absurd, with player names like Freemium Seraph and Mags Banananana and baseball-themed surreal horror elements like an enormous, vengeful peanut god.

However, jokes and short games aren’t enough to create a procedurally generated narrative that truly moves an audience. That requires careful attention to a randomness generator that is capable of telling stories.

To the Corner, You Fiend!

If random outputs were enough to get people excited, we wouldn’t have people writing stories at all, and just create fancier and fancier dice for people to stand around and goggle at. Chance-based narratives need to have room to create moments of special excitement or interest in order for audiences to center around those moments.

I wrote an article a few years ago on replay value that roughly covers this – in any game with a variable setup or gameplay, the random elements need to create meaningfully different outcomes. The most effective way to do this is to have a few “special” outcomes buried in a sea of otherwise unimportant noise. Cat Manning and Sam Cabo Ashwell talk about this in the context of Blaseball in depth in an article on Medium – Blaseball excites its audience because the truly special things only happen every so often, while most games are largely mundane affairs where the weirdest thing that happens is you might get a decimal point in your score. Chorby Short hitting 200 foul balls in the first game of a season is only interesting because there were so many other games and players who didn’t do anything that special.

You could boil this concept down to its barest essentials with bouncing screensavers that make it into the corner on rare occasions. If this screensaver always went into the corner, or travelled on a path that brought it into the corner about 50 percent of the time, it would be largely uninteresting – it’s both the promise of something out of the ordinary, something distinct, and the exceptionally rare fulfillment of that promise that keeps people so fixated on something so simple. The fact that it’s a corner is also important because corners are distinct – if the screensaver missed some middle inch of the bottom screen most of the time, nobody would notice it and nobody would really care.


Jelle’s and Blaseball are spectator sports, so they have the luxury of making their results 100% reliant on chance. As a designer of presumably interactive games, you can’t transplant the techniques that these games use one for one into your game, but you can draw lessons about meaningful chance-based outcomes that are useful in all but the dryest of Euros. How can you make random outcomes meaningful? What results can you make players strive for that might not even happen in a game?  How can you make a component that someone might root for? These questions could be the difference between a mediocre game and a great one.

The Illusion of Control, Part 2: Chrom Fireemblem Is A Dirty Cheater

Welcome back to my 2-part article series about various methods to disguise the variance in your game to give your players an illusion of control. You can read part 1 here.

There’s a very understandable revulsion to “being manipulated” that makes this subject a little strange to write about. After all, we can’t pop open the news without finding another grim headline about how our personalities and behaviors are largely the product of market forces that have been influencing us since we were children. The idea that a game, which we often play to reclaim a sense of control, is also “cheating” in a sense, can feel slimy and disheartening.

The key is that a well-designed game engages in manipulative behavior purely for the player’s benefit. A “fair” experience that doesn’t lie to or manipulate the player can often feel stark or brutally economical. This is especially true in video games, where nearly every game cheats, usually in the player’s favor, in order to better fit the model provided by human cognitive bias. The random number generator used by the Fire Emblem franchise is a nice, simple example: By using the mean of two RNG numbers to determine hit chances, it makes a 75% chance to hit “feel” more like how we think a 75% chance to hit should feel.

Just like how media is a safe way for audiences to experience emotions like fear and distrust that would be dangerous in the real world, it’s also a good space for creators to try psychological tricks that would be unethical outside of the game context. Of course, this applies only to techniques used within the “magic circle” of the game environment; manipulating your players into, for example, constantly paying for microtransactions in a mobile game falls back under the umbrella of questionable ethics.

Anyway, let’s look at some more techniques that I didn’t cover last time.

Chess: Disguise Variance Through Many Non-Variant Moves

Despite being a big-brain game for smart people and bad guys in action movies, chess and games like it have a fair amount of variance. They must have, because otherwise every single chess game between the same two high-skill players would turn out the same. However, chess has no “luck” per se beyond determining which player goes first; players are in total control of every piece at all time.

The trick is that players create variance through minute decisions that eventually compound into unknown game states. The average game of chess between high-level players is about 40 moves long; if you consider, for example, how moving a bishop two squares or three squares completely changes what pieces it threatens, you can see how each small move eventually creates a relatively novel gamespace. “Perfect information” Hobby games like Terra Mystica that use this form of variance go even further, with hundreds of individual decisions adding up to a game state that’s different even with identical setups, character selections, and players.

In a way, this method of creating variance isn’t an “illusion” of control – it is control. But it falls under the banner of this article series because it sneakily adds variance into a state where everyone involved is being perfectly rational. Instead of dice or a deck of cards, these games use the human mind as the ultimate random number generator.

Ra: Trick Players Into Logical Fallacies Through Weak Evidence

Humans are subject to several common cognitive biases. Many, many articles have been written about them, partially because they’re fun to write and partially because trying to rid yourself of them is a good step on the road to self-improvement.

But as game designers, we’re not out to force our players to change how they think; it’s better to go along with human nature, which includes human cognitive biases, and provide a game environment that works the way it “should.”

As an example, let’s look at the classic Reiner Knizia auction game Ra. In Ra, players draw tiles out of a bag and place them up for auction. Any tile taken out of the bag never goes back in; it’s either discarded at the end of the round or kept in a player’s tableau for end-game scoring.

This allows players to predict the outcome of future rounds based on previous rounds; for example, if you haven’t seen a lot of flood tiles, you might be more bullish on Nile tiles (which need flood tiles to score). If a lot of Ra tiles, which immediately start the auction, are drawn, you can be greedier and draw more tiles in hopes that you can bid on a better lot. These tiny edges might add up to an eventual win at the end of the game!

There is one problem, however: Everything in the last paragraph is a lie! There are so many tiles in the bag that the distribution isn’t substantially affected even if weird draws occur. To use the flood tile as an example: The bag has 180 tiles in total and 12 flood tiles. Let’s say you draw half (6) of them in the first round. If you drew 40 tiles total that round, in the first draw of the next round you’ll have a 6 in 140 chance to get a flood. That’s about 4 percent. The odds of drawing a flood tile as the first tile in the first round is 12 in 180, about 6 percent. So even though you drew half the flood tiles, the odds of getting more is changed only by an imperceptible amount that doesn’t really affect your strategy.

(Yes, I know I basically threw out a number of total tiles drawn at random, but 40 is my rough estimate for the number of tiles drawn in an average round. I’m more trying to illustrate my larger example.)

However, the odds have changed to a small degree, and enough of a degree to trick players into the gambler’s fallacy – even players who would normally know better. If you bid low on Nile tiles in a later round because you saw so many floods, you feel like you won because you knew there weren’t many left (ie. you weren’t due for one), even though the real reason was that the bag pulls worked the way you thought they’d work. Even better, if you do end up pulling more flood tiles, you can just blame the whims of fate for creating such a freak accident.

And you know what? This rules, because it works the way your mind wants it to work. Instead of fighting our brains’ incorrect perceptions all the time, we get to send them to Cognitive Bias Fantasy Camp, where we really were due for that flood tile. By using this method, games can have their cake and eat it too: We get variance, but we also get enough reason to believe that we can predict what happens next that we don’t feel like we lost because of luck.


As a designer, your responsibility is to make a game environment that’s fun for the player without being exploitative. Your responsibility isn’t to make a game environment where everything is as it seems. Just like how audiences accept that a stage magician is tricking them somehow, they’ll accept the same from your game, as long as you entertain them. I hope these tricks to disguising variance help you do so.