Below is the essay version of a talk I gave at Roguelike Celebration 2024. The talk itself is here, and the associated slides are here, although you shouldn’t need either to read the essay version below.

If you enjoy what you read, you can sign up to receive more like it here:

Thanks, everyone, for coming to Roguelike Celebration, and to this talk.

Introduction

This talk is all about combos!¹

First, we’re going to formalize a definition of comboness, centering on the notion that combos are, definitionally, non-obvious.
Next, we’ll go over the expanding self-improvement space available to the median person alive.
Then, we’ll go through some biographies of people who have excelled in their fields, looking at how their careers can be viewed through the lens of combos. We’ll also consider the notion that “high-comboness” paths to success will be increasingly common as the self-improvement space expands.
Finally, we’ll look at how the parallels between combos and the expanding space of real-world self-improvement can be used to achieve experiential goals in game design.

Special thanks

Special thanks go to Gareth Falkingham, Jon Gill, Emily Greer, and Scott Smitelli who provided review and discussion.

Part I: A Definition of Comboness

Most of us intuitively know what a combo is. It’s when you get two different upgrades in a game that happen to go well together, right? But let’s try to get as precise as possible and see where that precision takes us.

Let’s start with a first-draft formal definition of combos, and refine it until we get to something that agrees with our intuitions:

Given a set of three or more possible upgrades, a combo is:

A subset of upgrades,
of size > 1,
That has a value greater than the sum of its individual parts,
When accounting for conditions internal to the subset.

Let’s take the simplest possible example: A game with 3 upgrades:

A has value 2
B has value 1
C has value 0, unless B is also present, in which case it has value 3

Now let’s say you have to pick a subset of size 2 from these 3 upgrades. The values of each possible set are:

[A, B] = 3
[A, C] = 2
[B, C] = 4

This game probably isn’t fun unless you’re at the Candyland cognitive stage. But more importantly, this definition and example don’t actually reflect how the term is used colloquially by game players and designers. It leaves a loophole for cases such as:

A has value 10
B has value 1
C has value 0, unless B is also present, in which case it has value 3

Which leads us to:

[A, B] = 11
[A, C] = 10
[B, C] = 4

In which case [B, C] is more valuable than the sum of its individual parts, but isn’t a subset one would voluntarily choose, and therefore most people wouldn’t describe it as a combo. So let’s refine our definition to:

Given a set of three or more possible upgrades, a combo is:

A subset of upgrades,
of size > 1,
That has a value ~~greater than the sum of its individual parts~~ that exceeds the highest unconditional value of all subsets of the same size,
When accounting for conditions internal to the subset.

There’s still one important gap, though, between the common usage of the term “combo” and our examples, if not our definition. Our definition leaves open whether the conditional value is explicit or implicit; that is, whether there is a conditional, value-increasing effect directly flowing from the upgrades themselves, or whether the value increase occurs implicitly, after the effects of the upgrades in the subset are passed on into other game systems. Common usage of the term only includes the latter. So [B, C] would not be a combo if this were literally the text printed on the upgrades:

“A has value 2”
“B has value 1”
“C has value 0, unless B is also present, in which case it has value 3”

But if we changed the text on the upgrades to the equivalent:

“A has value 2”
“B has value 1, take a Foo token”
“C has value 0, and also grants 3 value for each Foo token”

More players of combo-centric games would now regard it as a combo. But, importantly, they would regard it as “not much of a combo,” because of how obvious it is.

In fact, the less obvious a combo is, the more “comboness” players are likely to attribute to it. This example from Magic: the Gathering that my CTO Gareth shared with me took five minutes to explain, and I think we can all agree (even without a live explanation) that it appears to have a lot of comboness.

This point gets to something that will be important to drawing out the implications of combos: “Combo” isn’t a binary property; it’s a matter of degree, and the matter of degree is dependent on the non-obviousness of the combo’s conditional value as much as on the conditional value itself. So let’s revisit our definition one more time:

Given a set of three or more possible upgrades, ~~a combo~~ the degree of comboness of:

A subset of upgrades,
of size > 1,

~~That has~~ Is, roughly, the product of:

The degree to which its value ~~that~~ exceeds the highest unconditional value of all subsets of the same size, and
The non-obviousness of the conditional value,

When accounting for conditions internal to the subset.

Or, if you’re bored of all of this logical formalization:

A combo is a subset of upgrades whose value is:

Higher than what you’d get if you picked purely on unconditional value, and
Non-obvious.

Part II: Life keeps getting more combo-licated

Now let’s talk about something (apparently) completely different: The complexity of modern life!

Figuring out your best life is more complicated for us than it was for, say, our hunter-gather ancestors. There are just a lot more options.

Now, this sounds like a zero-information platitude – “Things were much simpler back in the old days!” – but let’s try to put some actual structure on the complexity space of self-improvement offered to an individual over time. Obviously, we know that the best way to understand global, human-history-scale trends is with a shamelessly Stewart-Brand-esque diagram.²

Wow. That is…a lot of stuff. Look, some areas of self-improvement do go away over time; most people in the audience don’t have to worry about crop selection, or stabbing things with spears. However, the general trend is that individuals have to worry about more and more self-improvement areas over time.

And the number of areas people have to worry about isn’t the only thing that’s increasing. I’d argue that the complexity of the option space within each area has also been increasing over time. God forbid you be an incoming college freshman in 2024, in which case you’d have to answer most of what humans have had to answer throughout human history, along with all of the following:

Of the thousands of types of work you could do, which are going to have their wages and prestige rise or fall due to technological trends? What about the additional hundreds of types of work that might emerge in the next 50 years?
It is worth moving to an unbelievably expensive global-hub city just for some hazy notion of “connections” and “cultural development?” Which one? Wait, which ones support which types of work? Do you want to do those types of work?
The Internet has shown you an endlessly-shifting kaleidoscope of subcultures. Which ones are worth the time cost? The “I have to consume lots of media upfront to belong” cost? The weirdness cost? Which ones can you actually participate with in person in the city that you’re choosing?
How are you going to build an understanding of the contemporary digital-information landscape to avoid getting grifted, parasocial-ed, or addicted out of your money and time? Can you answer this question for every one of the dozens of digital platforms you have to interact with now, and the dozens of new ones you’ll encounter over the course of your life? Are the subcultures you’ve chosen making you more resistant to these threats? Less?

Now the important point here is that you can’t have it all. Investment in any of these areas is relatively expensive. If you’re a pre-industrial farmer and you want to learn about all the different things you can do to land to make it more productive–terracing, irrigation, green manuring, crop selection, etc.–even determining whether you’re good enough at it to justify further investment could cost you years of your life. If you’re our class of 2028 college freshman, and you want to figure out the usefulness of some deep skill area you’ve never tried, like quantum computing or pro-wrestling fandom or federal land-use regulations, the cost is just as high; but it’s even worse because you have so many more to choose from.

We live in an increasingly post-Malthusian world, so if you get this really, really wrong; you probably won’t die. But you could very well end up broke, alone, and miserable! Or even worse, you could end up…average.

So what’s going to become much more important in this exploding possibility space?

COMBOS.

Part III: Biography as Combo

Let’s go back to the short definition of combos we ended up with at the end of Part I:

A combo is a subset of upgrades whose value is:

Higher than what you’d get if you picked purely on unconditional value, and
Non-obvious.

Now let’s go over the biographies of three people who excelled in their fields by developing skills that not only didn’t obviously combo together, but weren’t even obviously useful in their chosen field. To put it another way, they acquired skills that had high conditional values but low unconditional values, and then managed to get the conditional values to kick in.

Biography #1: The Cratedigger

Let’s start with a humble person trying to break into the game industry.

This person has no experience. They didn’t get a degree in games, and they literally have no professional connections. But they’re determined.

Now, to call this person’s tastes in games un-commercial might be understating the case. Their favorite game is Ultima Online; and just so you know, we’re not telling a story about the 1990s. This person’s story starts in the early 2010s. For those who are unfamiliar, Ultima Online looks like this:

It is completely not obvious how this is of any contemporary advantage in making commercially successful games.

This person is also, as people used to say in the age of vinyl records, a cratedigger. Just like producers and DJs would rifle through discount crates in the backs of music stores, hoping to find an awesome-but-unknown track to spin or to sample, this person digs through obscure mobile games to find novel gameplay experiments. We’re getting warmer in terms of “maybe this is a skill you’d want to have to be a commercially successful game developer,” but plenty of people–of which there are certainly some in the audience of this talk–have pounded through thousands of obscure mobile games, weird hacked console ROMs, and ancient shovelware collections without spotting a single commercially-viable idea.

This person’s career in games goes so poorly that they end up working in fast-food to make ends meet. And then they do finally get a games job, but it’s…only sort of in games?It’s working on slot machines. Now, I have a big-tent view of what games are, and I (by default) regard mainstream casino games as a valid part of the broader human play experience. But among hipster indie developers–of which there are certainly none in the audience of this talk–the general perception is that the negative social value of working on casino games is one step short of leprosy.

But this person still doesn’t give up.

One day, they play a game that’s really, really fun. It’s got incredibly chaotic gameplay, but the game-design community doesn’t seem to be aware of it because it’s fairly new and only available on Android. What it seems to be missing is a really legible approach to its chaotic, large-scale battles, and a better approach to its progression system.

So what we have here is:

Passion for retro games with really weird (bad?) graphics
Video-game cratedigging
Experience making slot machines

And what we get is:

Luca Galante, creator of Vampire Survivors

Here are the non-obvious conditionals:

Ultima Online is a wildly visually-chaotic 2D, real-time game, using low-res assets from the late 90s. Many, many indie developers have adopted a pixel aesthetic because it’s usually cheaper to make assets for, and also you can get people who are nostalgic for the Super Nintendo Entertainment System to give you $15. The problem with retro-nostalgia game graphics is that they are normally fairly bad at clearly showing lots of entities acting in real-time, meaning that there’s extremely stiff competition in the few game genres that don’t have to worry about that, namely JRPGs and platformers. However, there is a now-mostly-dead tradition that knew how to do that…and it’s the tradition of 2D MMOs like Ultima Online.

Let’s actually take another at Ultima Online next to Vampire Survivors:

I don’t know for sure if the developers of 20XX and Rogue Legacy and Shovel Knight and Axiom Verge and Enter the Gungeon could have pulled off A) retro-console graphics while also B) having hundreds of active entities on screen, thus solving Steam top-of-funnel with SNES nostalgia while also breaking through into a new game genre to avoid gameplay competition. But nobody else actually did before Luca Galante!

Okay, but what about the cratedigging? Arbitraging game genres across platforms is one of the only ways to get alpha, in the investment sense, in the games business. In theory, indie developers should be looking for opportunities to import emerging genres to their target platform from other platforms, but in practice few do. In fact, I only know less than half a dozen people who are true cross-platform inspiration magpies, and half of those are in the audience of this talk right now.

The casino stuff is actually the easiest to explain.³ It’s not, as many developers salty about Galante’s success have claimed, something like “Oh the mass market is dumb, they just like randomness, it’s all gambling.” It is, again, an arbitrage opportunity. The proportion of attention that indie game creators pay to other platforms is out of step with the proportion of total human dollars that flow towards those platforms. The global gambling industry makes very roughly what the video games industry makes every year, and yet a miniscule number of game creators pay attention to it. There was going to be something worth importing if someone dug around; a point reinforced by the breakout success of Balatro this year.

Biography #2: The Ice Skater

All right, let’s try another one.

Let’s say that you’re an aspiring professional ice skater. And let’s say that you don’t make the Olympic team, and everyone tells you your skating career is over.

Your family wants you to join their pharmaceutical-distribution business. You’re aware of how it works, but you don’t particularly like distributing pharmaceuticals.

So you become a fashion journalist. You do that for about 20 years, but that doesn’t mean you’re necessarily prepared to do anything else. In your own words, “I didn’t know anything about dress design.”

And then you suddenly become a successful fashion designer–likely the best-known bridalwear designer on the planet–within the space of a few years.

I’m talking, of course, about statistically the least likely person to be mentioned in a game-development talk: Vera Wang.

I particularly love this example, because it turns out that there were non-obvious conditionals hiding at each step here:

Professional ice skaters don’t just learn routines, they also learn what it’s like to be wearing a complicated, expensive ceremonial outfit while a large group of people stares at them for an extended period of time. Like a bride.
Running any kind of distribution business isn’t just about merchandising and fulfillment; it’s about understanding margins, inventory complexity, and turnover. Most people who try to get into the fashion industry start, as far as I can tell, with jewelry or small leather goods like handbags, and end up running zero-customer, money-pit boutiques in the West Village or Venice Beach or Hayes Valley. When asked by the Harvard Business Review why she chose bridal as her beachhead market, Ms. Wang replied, “My father identified that as an opportunity. He didn’t work in the garment industry, but he was a businessman, and he saw that: bridal came with lower risks: It had low inventory, few fabrics at [a] time, and, since people will always want to get married, a steady stream of customers[.]”
Journalists don’t just learn how to write and edit, they also learn what gets coverage and what doesn’t. When was the last time you heard about a celebrity’s wedding dress where the design house was mentioned and it wasn’t by Vera Wang?

Biography #3: The military-intelligence archivist who invents shark repellent

All right, let’s do just one more.

Let’s say that you’re an archivist who works in military intelligence. You’re really, really good at filing, and people know it. You know where all the spies are.

Then, one day, you get assigned to a really weird project. Sharks are eating sailors; specifically your sailors. (If they were eating enemy sailors, presumably that would not be a problem.) Someone has to keep track of all of the chemical compounds that scientists are dumping into water to see if the sharks run away. You use your skill at keeping track of the spies to keep track of all of the different chemical mixtures that they’re trying, and precisely how they’re being blended. And the project works! You and the team invent a shark repellent that remains standard military issue for decades.

Then you get married and move to France. Most of your friends are fellow American expats. You’ve discover a small circle of people who are willing to dig into the culinary culture of your adopted country, but importantly, they’re not in any rush. They’re amateurs like you, and they’re willing to learn the basics from scratch, even if it takes a decade.

So what we have here is:

The ability to keep track of an enormous number of cards with similarly structured information
Experience managing shark-repellent R&D
A social circle enthusiastic about cultural arbitrage and willing to mess around as amateurs for years

And what we get is…Julia Child?!?!?!

The evidence that Julia Child was pursuing a non-obvious path was that she even looked like a bad bet to the experts. Houghton-Mifflin broke their contract with her for Mastering the Art of French Cooking because the manuscript she turned in seemed like too much of a reference work. Something that might have been written by a professional archivist, maybe? But it turned out she was onto something; according to her biographer:

[T]he publication of Mastering the Art of French Cooking instantaneously changed the entire American cookbook industry, leading more cookbook publishers to place emphasis on clarity and precision, and away from the “chatty and sometimes sketchy” style that had typified American cookbooks.⁴

Harbingers of a trend

So I’ll admit that I deliberately picked underdog stories here, but I picked underdog stories where the underdogs succeed not only through sports-movie-style grit–although all three of our subjects certainly had plenty of grit!–but because they acquired and exploited a non-obvious combination of skills.

There are certainly people who excel in their fields by following a straight and predictable path. Note that I didn’t pick anyone like Stephen Sondheim; if you’re already learning composition from Oscar Hammerstein as a young child, then your life might make for an interesting story, but not one that is interesting due to comboness.

But my contention is that if, as we explored in the previous section, the possible space of self-improvement is increasing at an increasing rate, then more and more success stories are going to look like these three: years of work acquiring apparently unrelated skills, followed by a “turn” where the skills might be coming together into something greater, but it’s going to take years of high-risk commitment to find out.

If more and more people are going to be living this way, or at least trying to, then more and more people are going to be feeling like the primary struggle of their life is searching rather than, well, struggling.

Part IV: What Combos Evoke

As noted at the beginning, this talk is not going to argue that games make you better at life. Sure, I’ve pointed out some ways in which life is sort of like games, but the point here is not that you should play games with combos to get better at your professional and personal endeavors. I’m pretty sure that Vera Wang has never played Magic: the Gathering, and I can guarantee you that Julia Child never played Vampire Survivors.⁵

So why did I bother going on and on about how real life has dynamics that sort of look like combos? It’s because I’m about to make an argument in the opposite direction: about how to make games evoke life.

Games, like any media work portraying a stream of events, work through a cycle of anxiogenesis and anxiolysis. This is a fancy way of saying that they create anxiety, and then they make it go away. You’re familiar with this notion if you’ve heard of rising tension and falling tension in literature, or if you’ve ever diagrammed out gameplay loops.

Now, if you want to be able to create anxiety and then take it away, you need a structural understanding of your audience’s anxieties. As a very simple example, I will now show you a page of text with “F – come see me” written on it:

Some of you are likely now having anxiety flashbacks. But if our hunter-gatherer ancestors saw this, they would, I would guess, have no reaction at all. Now I’m going to show all of you an image of a savannah during a drought:

Most of you would not have an immediate reaction of anxiety, but our hunter-gatherer ancestors might remember a particularly desperate dry season and consequently feel a deep-seated dread.

Importantly, if you want your work to have an impact on your audience, it must be strongly anxiogenic. It doesn’t matter if your intent is to force people to gaze unflinchingly at the unpleasant realities of life or to make the most narcotically escapist work possible;⁶ in the first case the irritant won’t be consumed without the relief, and in the second case, the sense of relief is in direct proportion to the irritant. A big pearl needs a big grain of sand. So if you want deep reactions of any sort out of your audience, you need to be able to evoke deeply.

As far back as we’re able to peer into the history of human storytelling, narrative has mostly involved a hero who has a relatively well-defined goal and just needs to stick to it. The archetypal hero’s journey is full of refusals of the call, obstacles, temptations, fearful challenges, etc. Structurally, it’s a single highway with some offramps, and the struggle is about not being weak or stupid or shortsighted enough to take the offramps. “Don’t chicken out and say no to God, Jonah!” “Don’t brag about how you’re better than the gods, Arachne!” And this made sense for a world where the primary economic virtue by a long shot was: “Just accept commonly-held virtues and then work hard at them.”

At no point during the archetypal hero’s journey does the hero sit down and say “Wait, my mentor didn’t give me a call to adventure, he instead gave me a list of possible calls, along with a list of possible resources I could acquire to make myself statistically more likely to achieve some calls, along with some vague hints that some combinations of resources are better than others when pursuing certain calls. So now what?”⁷

But, in our increasingly combo-centric world, this overwhelming tide of superficially-viable options is the core anxiety source for more and more people. If you want to create work that appeals to those people, you need to understand how to evoke those anxieties in them, and then how to evoke relief.

Our time here is far too limited to examine every possible design decision involved in upgrade systems and to also elucidate each one’s effects on anxiogenesis and anxiolysis. So we’re going to look at just one: constructed upgrades vs. drafted upgrades. Very broadly:

In constructed upgrade selection, the player has perfect knowledge of which upgrades they will be able to take over the course of the game, whereas,
In drafted upgrade selection, the player has imperfect knowledge of which upgrades they will be able to take.

We’re going to treat these two as existing on a continuum, although design reality is obviously more messy than that.

Now, in their pure forms, these two approaches to upgrades push the player toward a different relationship with combos:

Combo-wise, constructed tends to reward…		Combo-wise, drafting tends to reward…
Memorization Theorycrafting Reading “the meta” Defining optimization functions Finding global maxima	vs.	Improvisation Gauging probability Valuing optionality Calculating opportunity cost Finding local maxima

I’m going to make a Goldilocks argument here: neither strongly constructed nor strongly drafted are as anxiogenic–or as anxiolytic–as a balanced combination.

Think of the experience of engaging with a game in pure constructed mode. You’re an ARPG player reading the skills wiki, or a constructed-format CCG player poring over card databases. You’re very much engaged in a leisurely armchair exercise. You can read up as much about theorycrafting as you want, on your own timeline. Whenever you want to hop back into the game, it’s there.

Now think of the experience of a purely random upgrade system; something like the “let’s try to play with truly random blind decks” format every CCG group tries exactly once and then never again. You might have some anticipation about what you’ll get, but you’re not anxious that you might make a bad call. By taking decision responsibility away from you, the game makes itself less anxiogenic.

Now these two modes of engagement certainly have the ability to create anxiety. I’ve felt anxiety going into constructed CCG tournaments, and I’ve felt anxiety betting on games of pure chance. But these approaches are not going to be maximally effective at creating long-term audience engagement, because these forms of anxiogenesis are not structurally reflective of the anxiety people feel in their everyday life.

So what form is? Well, let’s look back at our three exemplar biographies. Within the subset of existing self-improvement areas that were opened up to each of them by chance, they had at least some control over how, and how much, to invest in each one. We’ll call the upgrade-design pattern that maps to this reality the “drafting mullet” approach: it’s drafting in the front, constructed in the back. To use Soren Johnson’s terminology of pre-luck and post-luck:⁸ the skill areas that are available at all are subject to pre-luck, and then player decision determines the allocation of upgrade points within them.

And, of course, games with this structure exist! The seemingly-omnipresent 3-pick-1 format of roguelite upgrade systems is popular because it’s a simple and elegant way of achieving the drafting mullet. As typically implemented, it’s only viable to pursue the subset of combos for which you are offered one component early enough in a run that you might still draw the others. For those combos that you actually are offered an option on, you have some freedom about how you deeply you dig into each one.

But there are plenty of other examples; I’ll leave it as an exercise for the reader how this principle is implemented in Heroes of Might & Magic, Caves of Qud, Rimworld, and autobattlers.

I’ll also leave it as an exercise for the reader to apply this theory of anxiogenesis, anxiolysis, and combos to all of the other possible elements in an upgrade system. Since we’re a roguelike-enjoying crowd here, I suggest the following design threads to pull on:

Methods of creating combo density
Rarity systems
The interaction of upgrade systems with simulationist world models

So here are your takeaways from this talk:

Combos are, definitionally, non-obvious.
Life is increasingly complex, and thus increasingly about finding combos.
If your goal is for your game to reflect the anxieties and reliefs of people’s real lives, design your game’s upgrade system to open a subset of possible upgrade areas, but then give players freedom to choose which combos within that subset to pursue.

Thank you, everyone! I’ll take questions in the breakout session.

The CCG notion of combo, not the versus-fighter notion of combo.
The following diagram is highly schematic; many of these activities are attested, in restricted ways and/or among elite social classes, to significantly before when I have portrayed them as starting.
Luca Galante has said in interviews that he didn’t learn anything about reward loops from working at a casino-games company, but he also switched from acknowledging Castlevania’s obvious influence on Vampire Survivors to being cagey about it, so I’m going with the obvious interpretation here.
Fitch, Noel Riley (1999). Appetite for Life: The Biography of Julia Child. New York: Anchor Books. p. 275.
Although I am unable to report whether Luca Galante has designed an evening gown, or made Coquilles St. Jacques À La Provençale.
Which is a morally acceptable goal in my book! As C.S. Lewis pointed out, the word for someone who opposes escapism is “jailer.”
It’s not like linear media has been completely falling down on attempting to create this type of anxiogenic-anxiolytic cycle. Complex supervillain plots, the techno-thriller genre, and the heist genre are all examples, though collectively limited in symbolic scope.
Soren informed me by email after giving this talk that “the terminology that has stuck for this distinction is ‘input’ and ‘output’ randomness…Not my preference, but can’t win them all!” I agree with Soren that the pre-/post- distinction is better, on grounds that it focuses the mind on subjective player experience rather than on the abstract operation of the game as a system. As discussed before on Seeing the Chessboard, terminological inaccuracy is just the pits.

Towards a Philosophy of Combos (Talk @ Roguelike Celebration 2024)