The corpus
The corpus was extracted by the distant-reader pipeline as Salter-style “flag but do not alter” Markdown: every diary's opening post is preserved verbatim while forum infrastructure is wrapped in YAML front-matter and <!-- artifact --> comments rather than deleted. This analysis stripped 245 thread-metadata blocks, 234 reaction blocks, 184 signatures and 114 banners before tokenising — plus a second pass that removed an unwrapped forum-navigation breadcrumb caught only at the phrase stage (see Verification).
Diaries per year
Top authors (handle · count)
Step 1 · Bag of words
Below is the complete bag of words — all 18,854 distinct tokens across the 503,576-token corpus, ranked by raw frequency, before any stopword removal. The first screen of twenty is almost entirely English function words (the, to, of, a, and); the first meaningful signal is the second-person address — you (7,111) and your (3,783) outrank nearly every content word, the native register of a developer diary perpetually addressing the player (“your realm”, “you can now…”). Page deep enough and you reach the long tail of hapax legomena — words like manzikert or pureblooded appearing exactly once. This function-word dominance is precisely what motivates the iterated stopword removal in Step 2.
| rank | token | count | share |
|---|
Step 2 · Stopwords, iterated
Four stopword waves, each an independent wordset. Toggle any combination on or off — and both tables update to the union of the active waves. Salter: “keep refining until you are happy with your results.” These toggles let you run the refinement yourself.
Survivors
| rank | token | count | share |
|---|
Removed by active waves
| rank | token | count | wave(s) |
|---|
“Removed” lists corpus tokens caught by the union of active waves; the wave(s) tag shows which active waves contain each term. With Group plurals on, regular plural and possessive forms are folded into a singular base that exists in the vocabulary (counts summed, every merged variant listed in the token cell); ≤2-character tokens are dropped by a separate length rule.
Step 3 · Word cloud
Rendered in your browser from the same #s2bag vocabulary as Step 2. Grey words belong to no thematic cluster. The default (all four waves removed, plurals separate) reproduces the cleaned cloud; toggle waves off to watch function words and dev-diary register flood back in.
Step 4 · Themes & genre
Genre of the corpus
This is mixed promotional–technical developer communication: equal parts design manifesto, feature reveal, and patch-note changelog, all in a second-person marketing voice. A meaningful fraction of the word-mass describes the act of building and patching the game (beta, feedback, fixed, mods, performance) rather than the medieval world — a fact the “Development & Community” theme makes visible and the critique returns to.
Characters & Personality
character, trait, stress, lifestyle, education, opinion, personality, portrait
The simulation's atomic unit. CK3 is a character RPG wearing a grand-strategy coat — traits, education, stress and personal opinion drive everything above them.
“The main design goals with Crusader Kings III were: Character Focus : Crusader Kings is clearly and unequivocally about individual characters, unlike our other games.” — Base game, dd_000_2019-10-24_dev-diary-0-the-vision.mdFaith & Religion
faith, religion, holy, church, rite, clergy, doctrine, crusade
Religion is modelled as configurable systems — tenets, doctrines, holy orders, holy wars — not as flavour. Three later expansions (By God Alone, Legacy of Persia, base reform) deepen it.
“A House Head has the power to legitimize bastards, call House members to war, and demand that they adhere to their Faith (refusal to convert will result in them creating a new Cadet Branch).” — Base game, dd_001_2019-10-31_dev-diary-1-dynasties-houses.mdRealm, Title & Governance
realm, title, vassal, liege, county, council, administrative, succession
The political skeleton: who holds what title under whom. Roads to Power's 'Administrative Government' is the corpus's biggest structural addition to this layer.
“Marrying in such a way that your Dynasty ends up on the throne of a foreign realm is therefore useful for other reasons than to just murder them until you inherit their lands.” — Base game, dd_001_2019-10-31_dev-diary-1-dynasties-houses.mdCulture & Peoples
culture, tradition, heritage, innovation, china, steppe, byzantine
Culture is a first-class system of traditions and innovations. The All Under Heaven sub-corpus pushes it furthest (China, Japan, the dynastic cycle).
“You will also be able to become a Merchant via a Decision when playing as a Duke or Count, as long as your Culture allows it.” — Silk & Silver, dd_003_2026-05-26_silk-silver-dev-diary-3-merchants-vision.mdWar & Conflict
war, army, siege, knight, levy, struggle, men-at-arms, casus belli
Present but notably subordinate: war is a tool characters reach for, framed through casus belli, holy wars and struggles rather than as the main event.
“A House Head has the power to legitimize bastards, call House members to war, and demand that they adhere to their Faith (refusal to convert will result in them creating a new Cadet Branch).” — Base game, dd_001_2019-10-31_dev-diary-1-dynasties-houses.mdCourt & Dynasty
court, dynasty, house, prestige, coronation, artifact, heir, grandeur
The Royal Court expansion (46 diaries) makes this the single most-discussed surface in the corpus — throne rooms, court positions, grandeur, artifacts.
“In the meantime, please enjoy Royal Court and will have more info for you in the weeks to come!” — Console Edition, dd_001_2022-03-01_ck-console-dd-1-vision-and-goals.mdEvents & Activities
event, activity, decision, scheme, tournament, travel, feast
The narrative delivery layer. Tours & Tournaments and Wandering Nobles turn the static map into travelling activities, feasts and journeys.
Map & World
map, province, terrain, region, steppe, iberia, peninsula
The spatial substrate. Map diaries cluster early (base game) and around region-specific packs (Iberia, the Steppe, Asia).
Economy & Trade
trade, merchant, goods, market, tax, gold, silk, herd
Long a thin layer, decisively thickened late by Silk & Silver (merchants, trade routes, markets) and Khans of the Steppe (herds, fertility).
Development & Community
beta, community, feedback, bug, fixed, mods, multiplayer, performance
The meta-register. A large share of word-mass is Paradox talking about making and patching the game — betas, changelogs, modding, performance — not about the medieval world itself.
Each expansion writes about its own thing
Most distinctive terms per expansion (mean TF-IDF within the sub-corpus). The DLC structure is legible in the vocabulary — every pack has a sharp lexical fingerprint.
| Expansion / line | Distinctive vocabulary |
|---|---|
| Royal Court | court · royal · culture · teaser |
| Base game | faith · realm · opinion · vassals |
| All Under Heaven | china · heaven · chinese · dynastic cycle |
| Tours & Tournaments | tournaments · tours · travel · rewards |
| Northern Lords | raids · war · score · expand |
| Fate of Iberia | iberia · struggle · peninsula · toledo |
| Roads to Power | administrative · byzantine · landless · adventurers |
| Khans of the Steppe | nomadic · herd · steppe · tributary |
| Legacy of Persia | iranian · persia · tax · clan |
| Wards and Wardens | wards · hostages · education · childhood |
| Legends of the Dead | plague · legend · legitimacy · measles |
| Wandering Nobles | wandering · travel · landless · lifestyle |
| Silk & Silver | merchant · goods · trade · markets |
| By God Alone | rite · archbishop · clergy · church |
| Coronations | coronation · oath · anointment · crowned |
Word-mass by expansion
Step 5 · Concept network
The network has a clear shape: a governance core (realm · title · vassal · liege · council · succession) sits at the centre, throwing out arms to the satellite systems — faith / clergy / crusade, war / army / siege, merchant / trade / gold, court / dynasty / prestige, and trait / stress / education. The two tightest peripheral pairs — merchant–trade and administrative–government — are also the two newest systems, suggesting recent expansions add self-contained subsystems rather than rewiring the core.
| concept | concept | co | PMI |
|---|---|---|---|
| merchant | trade | 29 | 3.16 |
| levy | tax | 18 | 2.76 |
| administrative | government | 54 | 2.52 |
| heritage | tradition | 12 | 2.32 |
| education | trait | 29 | 2.18 |
| education | holy | 12 | 1.98 |
| clergy | council | 5 | 1.97 |
| culture | heritage | 28 | 1.96 |
| army | siege | 31 | 1.95 |
| faith | religion | 39 | 1.94 |
| education | siege | 5 | 1.84 |
| province | siege | 14 | 1.80 |
| holy | war | 104 | 1.79 |
| clergy | faith | 19 | 1.72 |
Step 6 · Frequent phrases
Top bigrams
Top trigrams
The phrases read like a CK3 glossary: royal court, court position, holy order, casus belli, dynastic cycle, special buildings, throne room, silk road, declare war, cultural traditions, holy roman empire, taxes and levies. These are mechanics and proper systems, not marketing slogans — evidence the diaries are predominantly explanatory rather than hype.
Step 7 · Underwood verification
The three strongest findings — the top term, the top concept edges, and the top phrases — were each verified against representative corpus sentences. All passed.
“The main design goals with Crusader Kings III were: Character Focus : Crusader Kings is clearly and unequivocally about individual characters, unlike our other games.”Base game ·
dd_000_2019-10-24_dev-diary-0-the-vision.md“A House Head has the power to legitimize bastards, call House members to war, and demand that they adhere to their Faith (refusal to convert will result in them creating a new Cadet Branch).”Base game ·
dd_001_2019-10-31_dev-diary-1-dynasties-houses.md“In the meantime, please enjoy Royal Court and will have more info for you in the weeks to come!”Console Edition ·
dd_001_2022-03-01_ck-console-dd-1-vision-and-goals.mdEdge checks. holy–war: “…allows you to declare better Holy Wars…”. merchant–trade: “Trade is a new feature that allows merchants and republics to move trade goods…”. gold–prestige: “Merchants have a gameplay loop focused on making gold, which can be turned into prestige.” administrative–government: surfaces DD 148 “Administrative Government (Part I)”. No fabricated edges.
One flagged artifact. Caught and removed: an early phrase pass was dominated by identical strings appearing exactly once in 69 documents (“beta console suggestions”, “aars after action reports”). Grep traced these to an unwrapped forum-navigation breadcrumb — a row of sub-forum links — not authored prose. A line-level filter and a “count==document-frequency” rule removed it. This is exactly the failure the Underwood pass exists to catch.
Stance caveat. The high frequency of faith, holy and war means the diaries discuss and mechanise religion and conflict — not that the developers endorse them. CK3 turns belief and violence into configurable systems; the word counts measure design attention, not values.
Step 8 · Critique — what to trust, what to doubt
What the analysis gets right. The character-first thesis is not an artifact of counting — it is stated outright in Dev Diary #0 (“Crusader Kings is clearly and unequivocally about individual characters”) and confirmed by the term ranking, the network centre of gravity, and the phrase list. The DLC fingerprints are real and sharp. The relative subordination of war to faith, culture and court is a genuine and somewhat counter-intuitive signal for a game with “war” in its DNA.
What to be suspicious of. Frequency tracks how much Paradox wrote about a system, which is driven by release cadence and marketing priorities, not by how much play-time the system occupies. Royal Court tops the word-count partly because it shipped a flashy 3-D throne room worth many teasers. The corpus is also self-promotional: it will systematically over-state novelty (“new”, “now”, “added”) and under-report what is dull or broken.
What it cannot see. Dev diaries are not the game and not the players. This reading describes what the developers chose to say in public — not what the code does, not how the systems feel, and not what the community made of them (every reaction and reply was deliberately stripped to study the developer voice in isolation). Sarcasm, walked-back features, and quietly cut content are invisible to word counting.
Curation contingency
These findings describe this collection — 245 forum diaries indexed by the CK3 wiki — not “Crusader Kings III” as a whole and not the grand-strategy genre it samples from. A different sampling frame (patch notes only, or community AARs, or the wiki's mechanics pages) would surface a different vocabulary. The corpus determines the findings.