The corpus
Each reply post was captured via Claude-in-Chrome DOM extraction, quoted text split into artifact:quote spans and excluded from counts, and authors anonymised to stable pseudonyms. This is the community-voice half of the comparison whose other half is the developer-diary corpus.
Most active participants (replies)
| participant | replies |
|---|---|
| user_1c | 26 |
| user_xxu5uj | 15 |
| user_quj0zh | 9 |
| user_o0r468 | 8 |
| user_15a1 | 7 |
| user_wedhq3 | 7 |
| user_qvlcwb | 7 |
| user_wi66ec | 6 |
A long-tail conversation: 213 people for 436 posts means most participants reply once or twice. The thread is a crowd, not a debate among regulars — though a handful (top participant: 26 posts) carry sustained argument.
Step 1 · Bag of words
| # | token | count |
|---|---|---|
| 1 | the | 1,438 |
| 2 | and | 765 |
| 3 | that | 429 |
| 4 | you | 387 |
| 5 | have | 275 |
| 6 | for | 243 |
| 7 | not | 230 |
| 8 | they | 218 |
| 9 | are | 207 |
| 10 | but | 206 |
| 11 | would | 205 |
| 12 | more | 187 |
| 13 | with | 186 |
| 14 | this | 178 |
| 15 | like | 164 |
| 16 | was | 163 |
| 17 | your | 159 |
| 18 | army | 155 |
| 19 | will | 154 |
| 20 | were | 144 |
As in any English corpus the head is function words. Two things break through early: the developer's instructional register echoes back — you (387), your (159) — because players quote and address the dev; and the corpus' true subject announces itself at rank 18 with army (155), followed by siege, system, knights. The audience is answering the War diary on its own terms.
Step 2 · Stopwords & the cleaned vocabulary
| rank | term | count | share |
|---|
Step 3 · Word cloud
Step 4 · Themes
Register of the corpus
Where the developer diary defines systems in the second person, the responses evaluate, question and historicise them. The dominant moves are the feature request (“will there be…”, “please don't…”), the historical correction (pikes vs cavalry, what a medieval levy really was), and the comparison to CK2 / EU4 / Imperator. This is an expert, invested, sometimes pedantic audience.
Armies & Troops
army · armies · levies · troops · men-at-arms · retinue
The corpus' centre of gravity. Players interrogate the new Levy / Men-at-Arms split — what professional troops cost, whether levies are mere "meat shields", and how composition works.
“So armies don't have 3 commanders 1 for each flank anymore?”
Siege & Fortification
siege · castle · walls · besieger · trebuchet · engines
The single most-discussed mechanic. Replies dwell on siege weapons, fortified vs unfortified holdings, and the besieger-as-attacker rule.
“Will there be any restrictions on your character's actions if his main castle is under siege”
Knights & Commanders
knight · commander · martial · prestige · title
Knighthood and command draw intense debate — how powerful a knight should be relative to levies, and whether "Knight" is a title you bestow.
“I don't want to micro manage 30-40 minor titles and then these knights only have the power of 300-400 levies”
Tactics & Battle
battle · flank · tactics · advantage · attrition · supply
The loss of CK2's flanks/tactics phase is mourned repeatedly; supply and attrition as soft supply-lines are widely praised.
“Like in how army composition affected the tactics a flank used in CK2.”
Unit types & realism
cavalry · infantry · archer · pikemen · medieval · peasants
A strong historical-realism streak: long threads on whether pikes really countered cavalry, and whether medieval levies were conscripted peasants or trained freemen.
“Tactically speaking, the pike square hadn't been invented yet, and cavalry often used lances longer than the pikes of the infantry”
Comparison & meta
system · mechanic · ck2 · eu4 · imperator · warscore
Players constantly benchmark against CK2, EU4 and Imperator — the reflex of a sequel audience evaluating change against what they already know.
“I do kind of like the tactics system in CK2”
Step 5 · Concept network
The network splits into the same subsystems the diary introduced — a logistics arm (army–attrition–supply–siege), a battle-mechanics arm (battle–flank–tactics–commander–trait), the troop-type ladder (levy–infantry–cavalry–archer–retinue), and knighthood (knight–title–character). The tightest edge, attrition–supply (PMI 2.22), shows the community latching hardest onto the one genuinely new system: supply lines.
Step 6 · Frequent phrases
Top bigrams
Recurring trigrams
Only five trigrams recur in ≥3 posts — community prose is far less templated than the developer's patch-note register, where dozens of trigrams repeat. The reading itself surfaces a genre difference.
The phrases read like a wargamer's lexicon: men[-at-]arms, siege weapons, middle ages, light/heavy infantry, military service, peasant levies, shattered retreat, meat shields, naval combat. Authored discussion, not slogans.
Step 7 · Underwood verification
Each headline claim is grounded in a representative reply. All supported.
“Will there be any restrictions on your character's actions if his main castle is under siege”
“If my army has low supply and taking attrition, will the losses I take be proportionally dispersed among all "classes" of troops”
“And siege weapons make an appeareance too, I like it”
Stance caveat. A high count of siege or levies means the community discussed them most — driven by what the diary chose to reveal — not that these are the game's most important systems. The reading measures attention, not value.
Step 8 · Critique — developer voice vs community voice
What the reading shows. Run beside the developer-diary corpus, the contrast is sharp and quantified (log-likelihood keyness): the developer over-uses precise second-person mechanics — you, your, advantage, troops, pursuit, progress, limit, damage — the register of someone defining a system. The community over-uses evaluation and speculation — like, knight, nice, would, think, yes, commanders — the register of people reacting to one. Developers specify; players appraise, historicise and compare.
What to be suspicious of. This is a single thread — the responses to one diary (War), so its vocabulary is war-shaped. The community's apparent priorities reflect what this diary surfaced, not the game as a whole. And forum repliers are the most invested slice of the audience; silent players leave no text.
What it cannot see. Tone — sarcasm, excitement, exasperation — is largely invisible to word counts. The frequent comparisons to CK2/EU4 carry an affect (nostalgia, anxiety about change) the frequencies only hint at. And every reaction emoji and up-vote was deliberately excluded to isolate the written voice.
Curation contingency
These findings describe the reply thread to CK3 Dev Diary #3 — 436 posts — not the CK3 community at large. A different diary (a faith or culture diary) would surface an entirely different community vocabulary. Scrape the remaining threads and this becomes a corpus-wide portrait of the audience; for now it is one loud, knowledgeable room.