Civilization Revolution — The City-Builder That Plays as a Rush

CivRev was designed as a fast, friendly Civ — a city-builder for the console. Competitive/1v1 play decoded it into something else: an opening decapitation rush feeding a wide-expansion MIRR snowball that ends the game before its intended midgame. The intended identity (build tall, develop cities, win on culture/economy/science over time) turns out to be a stack of dead verbs. This is specimen #2 for the Dead-Verb Test and the cleanest live confirmation of the MIRR 4X thesis the vault has.

Links: The MIRR 4X Framework (the hurdle-rate engine), The Dead-Verb Test (most options dominated), The Hollow Opponent (the deterministic AI exploited), Capability Without Leverage (liquidate to gold), The Dominance-Frontier Lens (the shrinking frontier), Randomness as Termination (N≥3) (why N=2 map-RNG is the wrong place) Project: Civilization Revolution — Project Hub (the working home for the data + the engines/generators that explore this space). Grounding: static data captured Tier-1 (community wikis) in raw/civrev-civilopedia-tables.md, raw/civrev-civilopedia-great-people-relics.md, raw/civrev-terrain.md, raw/civrev-combat-and-mechanics.md, raw/civrev-resources.md. The hidden formulas (city growth, governments, AI tempo, combat win-%) are flagged below as open gaps (§6); no good measurement oracle is currently in hand (the original no longer runs after the iOS 32-bit cull; CivRev2 is a content reskin with a too-limited tablet UI). Source = competitive/1v1 CivRev play + Chris’s analysis; numbers cited are confirmed-from-tables unless marked (open).

1. The opening: an immune scout + a deterministic AI = free capitals

Above the easiest difficulty you start as a Settler (0/0/2), not a city. Two exploits stack:

The immune scout (a ruleset gap): you can’t attack a civ you haven’t contacted, and the lone settler doesn’t trigger contact — so a move-2 unit roams enemy ground untouchable, banking hut/relic gold and scouting AI capitals.
The hollow, deterministic AI (Hollow Opponent): it always develops the same way — ~5 turns to its first unit, which then wanders off to explore, ~10 to the next. The capital is undefended on a knowable clock, and the human knows the script while the AI re-rolls blind every game (the best-response-vs-minimax asymmetry).

Found a city two tiles off an AI capital, put both citizens on forest (+2 prod each → 4/turn), pop a Warrior (cost 10) in ~3 turns or rush-buy it with hut gold, and walk into the empty capital. Good map RNG → 2–3 capitals captured before the game “starts.” (Open, needs a measurement oracle — see §6: does a settler really avoid contact; exact AI tempo; does a lone Warrior take an empty capital or do capitals have inherent defense.)

2. Gold is the master resource — because it’s liquid tempo

Gold rushes units (~2 gold/hammer; a Warrior ≈ 20 gold vs ~2.5 turns). By MIRR, the opening reinvestment rate is the highest it will ever be — taking a capital ~doubles you — so liquid, instantly-deployable capital beats illiquid build-queue production. Everything converts into the gold→rush loop:

huts, barbarian kills, and named/resource tiles pay gold;
a junk hut unit (Spy 25, Caravan 30) you won’t use is capability without leverage — sell it for gold and it becomes a capital-taking Warrior;
a Galley (30, move-2) is best as an explorer/relic-finder; Knights Templar → a free Knight (4/2/2) skips all of Feudalism and roams dominating Warriors (atk 1) and Archers (def 2); a favorable early Wonder can win outright in 1v1.

3. The main theory: wide beats tall on the hurdle rate

Chris’s words: “building stuff, even though it shows positive ROI, does not meet MIRR.” That is the MIRR thesis. The hurdle isn’t zero — it’s the empire’s own growth rate, which going wide sets very high:

The Republic settler exploit: at the low gate of Code of Laws (60) you can run Republic and build settlers for 1 pop. A 3-pop city → settler (−1) = back to 2, plus a new 3-pop city: net +2 pop and +1 city per cycle (Republic 1-pop rule + founding-pop = measure). Minting a city is far cheaper than growing one organically (~2→3 pop ≈ 20 food — measure). Food is dominated.
The multiplicative snowball: the first-to-discover bonuses are all “+1 [yield] in every city” (Literacy +1 sci/60, Irrigation +1 pop/60, Engineering +1 prod/130). Value = bonus × city-count, and city-count grows geometrically — so per-city bonus × geometric cities = super-linear science with no braking term. Result: tech victories before 0 BC.
The tech tree is a thin spine, not a tree. Most techs are dominated; you want the per-city-bonus techs, Republic, and key wonder-unlockers. And backfill breaks the DAG — Great Library (techs ≥2 civs hold), Lost City of Atlantis (several free), Oxford (an advanced tech), Apollo (all of them) let you skip prerequisites and grab them free. The designed pacing is routed around.
- Tool-confirmed (tools/tech-frontier.py, a deterministic generator over the captured table): just 4 of 46 techs sit on the cost-vs-value frontier (Masonry, Irrigation, Railroad, Industrialization) — the spine is even thinner than it looks. And the rush-order flips at 1→2 cities: one-time bonuses (Free Walls) lead the single-city opening, per-city bonuses (Irrigation, Literacy) take over the instant you expand — rush first, then go wide. Separately tools/trap-detector.py (unit stats × the tech DAG, no soft params) flags Navigation as the lone tech-trap: a dead-end leaf whose Galleon is outclassed by Steam Power’s Cruiser — a tech you’d research anyway. Normal progression is not flagged; only the off-spine detour is.

4. The dominance cascade — what dies

Yields: hammers/trade > food > culture. You mint pop instead of growing it (food weak); culture’s payoff needs the tall play that’s already dominated (culture dead).
Great People pass through MIRR too — “gold now vs 10× by endgame” is a discount calc. The only uses that clear the hurdle are tempo on contested spikes (Great Scientist → a key tech now; Great Builder → win a wonder race now) and per-city effects scaled by width (+1 pop/city × wide). GP as a goal fails the hurdle — they’re cheap enough to get incidentally.
The intended game. Build-tall, culture, slow economy, and most of the command surface are dead verbs. The deepest finding: the game’s designed identity — the city-builder — is itself the biggest dead verb.

5. 1v1 makes it worse — the frontier collapses

The dominance-frontier verdict on design quality is the size of the viable-strategy set, and in 1v1 it shrinks to ~one line. Map RNG becomes instant death: this is randomness in the wrong place — in N≥3 it’s a healthy termination mechanism against the gang-up equilibrium, but in N=2 with a dominant strategy there’s no coalition to break, so map variance just decides who executes the one line first — a coinflip, not agency.

6. What’s grounded vs. what’s still open

Grounded from the captured tables: settler 0/0/2, Warrior 10, forest +2, Knight 4/2/2, all first-discover bonuses, wonders, backfill mechanics, the resource→tech map (+ Game = +3 food). Still open (load-bearing, unmeasured): Republic 1-pop-settler rule + founding pop; the city growth/food curve (turns wide-vs-tall from “tends to beat” into “beats by X”); AI opening tempo; the rush-cost formula; unit sell values; settler contact-immunity; per-resource bonuses (paste the 403’d Fandom table). No measurement oracle is in hand: the original CivRev no longer runs (iOS 32-bit cull), and CivRev2 is a content reskin (same mechanics, so nothing new to compare) with a tablet UI too weak to experiment in — no unit-selling, auto-advancing turns, no city-worker assignment (the build-vs-tech/gold lever the thesis turns on). A handheld ROM (the DS version) would be the better, more RE-accessible oracle, but it’s unacquired and low priority. So these magnitudes stay parked; the structural claims above stand on the captured tables regardless.