Civilization II: The Loop That Stole Decades

In 2012 a player calling himself Lycerius posted about a game of Civilization II he had been playing, on and off, for ten years. It had reached the year 3991 AD. Three superpowers remained, locked in a war that had been going for two millennia of game time; the ice caps had melted so many times that most of the planet was swamp; every city was a fortress under a permanent state of emergency, and nothing had changed in eight hundred turns. It became one of the most widely circulated stories the medium has produced, and it is usually told as a curiosity.

It is better read as a lab report. Lycerius had accidentally run Civilization II’s core loop to termination and documented what the machine does when the content runs out and the loop keeps going. The answer is that it keeps going. That is the whole design.

The turn is not the loop

Civilization II arrived in February 1996 from MicroProse, led by Brian Reynolds with Sid Meier in an advisory role, five years after the original. The obvious changes are cosmetic and structural in the ordinary way: the map went isometric, combat gained hit points and firepower so a fight became a series of exchanges rather than a single roll, Fundamentalism showed up as a government that traded science for fanatics, and there were full-motion advisors in a High Council who shouted at you in period costume. All fine. None of it explains the ten years.

The thing that explains the ten years is that Civ II never presents you with a stopping point, and it achieves this without any deliberate trickery at all.

Consider what is pending at an arbitrary mid-game moment. A settler reaches its destination in three turns. A city grows to the next population point in five. A wonder completes in eight. A technology lands in twelve. A trireme reaches unexplored coast in two. A war you started resolves in some indeterminate window around twenty. Each of those has its own period, and the periods are unrelated to each other — they are set by production values, food boxes, beaker totals and movement rates that were computed independently.

So the timers never align. At literally any moment you care to stop, at least one of them is one or two turns from paying out, and the cost of finding out is a keypress. And here is the closing of the trap: every completion spawns new timers. The settler founds a city, which starts a growth clock and a build clock. The technology unlocks three more technologies with their own periods, and a unit that makes you reconsider the war. The wonder finishes and you pick another. The braid regenerates faster than it resolves, permanently, by construction.

That is the entire mechanism. No variable-ratio reward schedule, no daily quest, no engagement metric — the industry would not have the vocabulary for another decade. Just a set of unsynchronised countdowns in a system where finishing one starts two more. Reynolds’s team built the most effective compulsion loop of the century out of arithmetic that nobody was thinking about in those terms.

You can test the claim negatively. Games with designed exits — a level, a chapter, a mission — have alignment points where every clock reads zero at once, and that is precisely where people put the controller down. X-COM has them: the Skyranger comes home, the mission scores, and there is a clean seam. Civ II has no seam anywhere between 4000 BC and the heat death of the save file.

Why the combat model is wrong, and why it matters less than you think

The famous complaint is real. Civ II’s unit strengths are flat numbers with no era scaling, so a veteran phalanx — defence 2, invented around 3000 BC — fortified in a city on a hill behind walls can and does destroy a battleship with attack 12. The multipliers stack until an ancient spearman is arithmetically superior to a dreadnought, and the game reports it with a straight face.

This produced thirty years of jokes and one genuinely bad design consequence: because defence scales through terrain and fortification rather than through technology, the optimal military posture across most of the game is to sit still. Reynolds’s hit-point model made combat longer without making it more legible. Civ IV eventually fixed the underlying problem by making obsolete units cheap rather than merely weak, and Civ V’s move to hexes in 2010 finally attacked the stacking that made the phalanx invulnerable in the first place.

The other flaw is worse and less discussed. Infinite City Sprawl — founding cities every two tiles, as many as physically fit, each one small and each one a net positive — is the dominant strategy, and it is dominant because city output has no meaningful penalty for density. The optimal way to play Civilization II is to cover the earth in villages and never build a big city, which is a direct contradiction of the fantasy the game is selling. Every Civ since has spent design effort on making sprawl hurt, and that is a two-decade argument with a 1996 bug.

Both flaws sit inside a game people played for ten years. The loop does not care whether the combat model is sound. That is a slightly frightening thing to learn about design, and I think it is Civ II’s most important lesson: engagement and quality are separable, and a broken system inside a good braid will be played to death anyway.

The real ancestor

Meier has been open that Civ’s ancestor is Empire, Walter Bright’s mainframe game from the late seventies, and that SimCity in 1989 was the proof that a game with no win condition could sell. Meier’s own Railroad Tycoon had established the house grammar a year before Civ shipped — a map, an economy, and a clock — and Pirates! had already proved he would rather give a player a sandbox than a plot.

The deeper ancestor of the timer braid is the board wargame’s impulse structure, and Civ’s genuine invention is to have removed the human opponent’s patience from the equation. Around a table, the natural stopping point is somebody’s bus. On a PC, at half past one, the only thing between you and 4 AM is whether some clock reads zero — and it never does.

Downstream, the honest verdict is that the best game in this lineage was made by the man who made this one, three years later. Alpha Centauri in 1999 is Reynolds at Firaxis running the same braid with a coherent setting, a genuinely strange tech tree and writing that treats its factions as arguments. Civ II is the template. Alpha Centauri is what the template was for. If you have twenty hours and a choice, I would send you there — and I would send you to Master of Orion 2 for the same era’s best expression of what a research economy can do to a player’s imagination.

Where it stands

Civilization II is a game with a broken combat system, a dominant degenerate strategy, a late game of unit-shuffling tedium, and one of the two or three most consequential loops ever assembled. All of those statements are true at once, and the loop wins, which tells you something about what players actually buy.

Play it on PC — Test of Time and the Multiplayer Gold Edition both still run, and Freeciv has been a free, maintained reimplementation of the idea for over twenty years. Play it on a night when you have nothing on in the morning, and understand that the game has no opinion about your morning.

Spoilers below

There is little to spoil in the ordinary sense, so take this as endgame notes.

Civ II’s default settings end the game in 2020 AD. If nobody has conquered the world or launched a spaceship to Alpha Centauri by then, it scores you, ranks you against a table of historical leaders, and stops. That ranking screen is the only authored moment of judgement in the entire design, and it is deeply anticlimactic by intent — you get told you played like some emperor or other, and the throne room you have been decorating for six hours does not come up.

The spaceship win is the more interesting one, because it is the only place the braid resolves. You build structural and component and module parts across many cities, launch, and then wait — a fixed number of turns while the ship crosses space and the game continues around you and other civilisations can, in principle, capture your capital and destroy the ship in flight. It is the one clock in the game that finishing does not spawn another clock. It ends.

Which brings it back to Lycerius, who simply declined the 2020 stop and kept pressing the key. Ten years and two thousand in-game years later, the world was ice-melted swamp, three empires held a permanent front line, and his cities ran Fundamentalism forever because democracy could not survive the war weariness of a war that could not end. Nothing in Civilization II broke to produce that. Every system was working exactly as specified. The braid regenerates, so the game continued, so he continued, and the report he filed is the most complete piece of criticism the game has ever received — written by accident, over a decade, by someone who thought he was just taking one more turn.