The Networks of Cultist Simulator

Cultist Simulator reveals its world through fragments and its lore is discovered by considering these fragments in relation to each other. The game’s Discord server has an active community that discusses the lore of the world (now known as the Secret Histories), and there is plenty of room for interpretation about what is going on. This article will use the tools of network analysis to get a high level overview of Cultist Simulator, how the game works, and gain insight into its world. This kind of analysis is just another point in a broader discussion, but it can provoke new trains of thought and offer a unique perspective on a game that is otherwise focused on navigating its world with limited information.

Before proceeding with the analysis, I owe some people a few words of thanks. Lyrositor was remarkably patient in addressing my questions and his work on the Frangiclave made the whole project possible. Lottie and Haley’s (from the special thanks) warm response to an early version of this project motivated me to move forward with it. Finally, Lottie and Alexis have been remarkably supportive of my Cultist projects in general, on top of building and maintaining a game that was worth putting all this effort into in the first place.

Mapping the Mansus

While this article is concerned with the results of the analysis, it is useful to understand what is being measured and why. A network consists of individual actors, called nodes, and the relationships between nodes are called edges (no affiliation). The network for Cultist Simulator was built using the following rules:

  1. All of the elements (usually cards) are included as nodes unless they are used for debugging.
  2. Edges are limited to direct references or inclusion in recipes (“consume X to get Y”) and are weighted by the values they take in game (for example, Winter lore with a value of 2, will get a weight of 2).

These two rules appear simple but carry some counterintuitive implications. Rule 1 means that there are multiple Neville nodes since Neville, an acquaintance is a different element than Neville, a Key. This can be considered the “more than one history” rule that allows us to go from a worm’s eye view to a bird’s eye view of the game. Rule 2 may be more contentious. Since the network only contains direct relationships, texts like the Ceaseless Tantra have no direct relationship with an aspect (in this case Heart) since it is the lore contained in the text that has the corresponding element. The strictness of rule 2 prevents odd circumstances like guardians and woulds becoming unusually well connected, or summons getting an edge with everyone they can eat.

One additional guideline for the Cultist Simulator network is that it only considers outcomes that are certain. This means that translating a book will create an edge between the translated and untranslated book, but the nightmare Connie Lee who has grown stronger with each failed assassination attempt is nowhere in the graph since her additional protections are random. This additional guideline does not have the strength of a rule because I have made an exception for the Mansus to treat it a bit more like locations and expeditions.

The Network

The network representation of Cultist Simulator from the rules above has 4,081 edges between 1,310 nodes. To put this in perspective, a core set for Magic: The Gathering contains an average of 311 cards, and an expansion cycle (usually 3 sets) would, on average, contain 578 cards. The diameter of the graph is 10, meaning the longest path to any given (reachable) node is 10 edges. For illustration, a graph of the Royal Canadian Navy with edges between the next rank would have a diameter of 10 (the distance from Naval Cadet to Admiral). Here is what Cultist Simulator looks like

Caution: This file is huge but beautiful

With 1,310 nodes there are 857,395 possible edges. The ratio of actual edges and possible edges is known as the density, with 1 being a graph where every node is connected to everything else. The density of the Cultist Simulator network is approximately 0.005, meaning it is sparse. The sparseness of Cultist Simulator reflects the fact that the player is thrown into an interesting world with only a few threads to pull on. Early backers of the game may also recall a first hand encounter with the sparseness of the game since there were no restrictions as to what cards could be placed into slots. The release version of the game informs players about what kinds of cards can be placed into a given slot, eliminating a certain number of choices and reducing the number of potential relationships a player will reasonably consider.

Sparseness is helpful in identifying communities in a network. Communities can be identified by looking for dense connections within a group and sparse connections to nodes outside of it. The communities that emerged from the analysis can be classified as follows:

  • Money
  • 8 Principles (all but Secret Histories. More on this later)
  • Patrons
  • Mortals
  • Seasons
  • Desires
  • Abilities
  • Long
  • Auctionable
  • Societies

Here are some of these communities close up.

The Mansus contains the Principles. The communities are coloured with their respective Principle.
A very large community in the mundane world: things you can sell.
Mortals are another large community. They may be part of the ordinary world, but hints of colour show how close they are to the Mansus.

These communities line up well with the aspects in the game. However, some of the communities that are identified might seem better suited as subsets of others. For instance, analyzing the network draws a distinction between mortals and patrons. Sharp eyed players may notice this relationship in the game on its own, but patrons’ loose relationship with mortality was clear enough in the structure of the game to get a community on their own without altering the analysis to enforce this relationship. In contrast, the mundane world (denoted money in the list above) does not spin out any interesting communities.

Data driven analysis can become interesting when it seems to go wrong. Secret Histories have their own set of lore fragments, a cult, tools, and articles, and yet the network identifies all of its nodes as members of the Lantern community. The status of Secret Histories is something of an open question among fans of the lore. A deeper dive into the unknown complexities of the world and its many pasts revealed this comment from Alexis. The analysis reveals that the Secret Histories’ ambiguous status is embedded into the structure of the game. Network analysis will not definitively answer this question, but it does provide additional evidence. Secret Histories are not structured like any of the other Principles, but Principles are not necessarily defined by their sparseness in networks.

Which Principle is best?

Network analysis can provide information about individual nodes as well. With 1,310 nodes to choose from, it is helpful to start with an interesting question. Most ambitions can be satisfied with any Principle and so one of the earliest and most important questions an aspiring cult leader must face is what Principle to dedicate themselves to or, to put it in simpler terms, which Principle is the best? But what do we mean by the best? Since the network graph concerns itself with relationships between elements in Cultist Simulator, we will focus on the most influential elements.

Influence can take many forms. One simple example would be a person with lots of followers on a social media site. Another might be a hypothetical friend who also happens to be friends with the Prime Minister. This friend may not have a large following on social media, but they are very influential since they can form an important link to a community (in this case, the government) that would otherwise be inaccessible. Network analysis uses a measure known as centrality to quantify influence. We will consider three measures of centrality:

  1. Degree Centrality: The number of nodes that directly link to you (or, equivalently, the number of edges between the node and everything else). This is the “how many followers do I have” measure.
  2. Weighted Degree Centrality: Similar to Degree Centrality except that it considers the weighting of each edge. For Cultist Simulator, this accounts for the intensity of the connection and means that an edge for a high value lore counts for more than low value lore.
  3. Betweenness Centrality: The number of times a node is on the shortest path between two other nodes. You might consider this a measure of a node’s ability to be a gatekeeper. Cultist examples will follow, but for now consider that Romeo and Juliet doesn’t move past the first act unless Mercutio has a high betweenness centrality.

Here are the top 15 elements by each measure:

DegreeWeighted DegreeBetweenness
AuctionableAuctionableAuctionable
MortalLanternMortal
TextEdgeDread
FollowerGrailJob
DiscipleHeartGrail
LanternMothFunds
SuspiciousForgeText
IndependentWinterLantern
EdgeKnockEdge
GrailMortalMoth
RivalmarksRivalmarksLocation
HeartTextHeart
LoreFundsProject
JobFollowerDesire
ToolSecrethistoriesApostlegrail

Principles are less influential than expected when measured with degree centrality. Lantern is the most influential by this measure, but fails to enter the top 5 and only four (Lantern, Edge, Grail and Heart) make it into the top 15. Auctionable’s position in the top reflects the fact that many elements in Cultist Simulator can be sold, and, of course, many of those objects are texts which find themselves in position three using this measure of centrality. Considering that degree measures the number of links (edges) a given element (node) has, the ranking by degree is less surprising. The Mansus is a special place that is only accessible to those with particular knowledge, which is not a trait one associates with lots of connections to all sorts of actors. The Principles do manifest themselves in the world through reason and hired goons (Moth’s absence from the top 15 is notable in this case given its association with passion), but the elements with the largest number of connections are the sorts of things we pass by in every day life: objects and people.

Using weighted degree centrality improves the overall ranking of the Principles, since their edges often have significantly higher weightings than what is found in the mundane world. Neither measure is bad provided it is appropriate to the question we are trying to answer. Cultist Simulator players, much like the character they play, are more likely to be interested in the occult world, and while this world may not have lots of connections (which is what makes it distinct from the mundane world), their importance is reflected in their weightings. Weighted degree centrality reflects this relative importance by being the only measure where all principles are present in the Top 15 list, with most getting high rankings. The ordinal ranking among the Principles is the same as for degree centrality, since most Principles are adjusted by the same weights (the increasing intensity of each lore card). As with the unweighted ranks, the ranking is also likely influenced by how the Principles manifest themselves in the world. Edge’s relatively high ranking is a result of how many mortals are associated with it. Lantern, Heart, and Moth have a similar argument given their association with the stats in the game: reason, health, and passion. Overall, weighted degree centrality acknowledges the importance of the Principles and their ordering reflects which Principles manifest in the world more readily.

Ranking by betweenness centrality offers a shift of perspective with results that may be less obvious but make sense on reflection. The two degree rankings were different ways of measuring the same type of influence, while the betweenness ranking is more concerned about optimal positioning in the network. When considering influence as the element that is most likely to get you where you need to go, Lantern must give way to Grail (Ezeem approves), reflecting its specialization in desires and how to fulfill them. Funds move up substantially in the rankings, consistent with its role both in the game and real life as a medium of exchange (a similar case is the high placement of job and made explicit in its description as “an arrangement to exchange one’s life for money.”). Even dread has its uses. When considering a high betweenness as the ability to get from any one point to another efficiently, most of the entries in this column makes sense. Locations have the things we want (Morland’s, for instance, provides access to a large number of texts), texts grand lore which grant access to the Principles and everything they touch. High betweenness elements are the things you use to get things done in the game.

Centrality is not limited to the three measures used here and, as we have seen above, the choice of measure can affect which elements get a higher ranking (though the Principles are strikingly consistent across measures). When considering a broader range of centrality measures and counting the frequency a Principle appears in a given position, the Principles have the following ranking:

  1. Lantern
  2. Grail
  3. Edge
  4. Heart
  5. Moth
  6. Forge
  7. Knock
  8. Winter
  9. Secret Histories

Lantern is best, Winter could probably talk a bit more, and Knock is also hampered by pausing before it speaks.

What is any of this good for?

Just as Cultist Simulator is perfectly enjoyable without jumping online and reading all the discussion about the lore, its lore is perfectly enjoyable without bringing the social sciences into things. Data may be useful in sharpening questions and inspiring discussion, but this kind of analysis will only ever be useful in so far as it enhances one’s enjoyment of the game.

The main enjoyment I have gotten by doing this kind of investigation is to see how well the lore of the game is reflected in its structure, and how certain real life concepts map quite well to the way they are portrayed in the game. I like that Grail, the Principle associated with wants and their satisfaction has a high betweenness centrality, since these seem like they should go together. The game would be just as fun if the funds card didn’t have a relatively high betweenness centrality, but the fact that money works the way much as it does in the real world is a nice little validation of the exercise. Even seeing little isolated islands of DLC related elements gives a glimpse into how the game works.

As a long time fan of Lantern, I am naturally happy to see it place so well in the rankings, but the take away is more than just an opportunity for smugness. Ultimately, I think the ranking reflects the fact that Cultist Simulator is something of a Lantern game. It’s all about reading tomes, figuring out the game as it goes, and having big flashes of revelation. This partly reflects my own interests (and isn’t intended to downplay passion and creativity), but I also think that a lot of what the player experiences and does when playing is very Lanterny. Lantern’s high placement in the centrality rankings could reflect an intention, or something going on in the background while the game was being developed, or could be a pure and happy coincidence. Whatever was really the case, Cultist Simulator is, to me, a Lantern game, and I like it because it’s a Lantern game. A network analysis of Cultist Simulator is just a way to apply some of my real life Lantern skills to the game and be delighted by how well many of the game’s concepts are reflected in its structure. In the end the analysis should be a way to enhance our enjoyment of something we already like. At least until The Exile comes out and turns all this on its head.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s