X-COM: UFO Defense — The Tension Machine
Julian Gollop built a strategy game out of irreversible loss and things you cannot see, and nobody has improved on the arithmetic

Contents
The most frightening thing in X-COM is a door.
Not what is behind it — you have no idea what is behind it, that is rather the condition of the entire game — but the door itself, and specifically the fact that opening it costs time units. Your rookie has eighteen left. Opening the door costs four. Stepping through costs another four. A snap shot costs roughly a third of his total. Do the sum and you find you can open the door and look, or open the door and shoot, and the sum is the game. Mythos Games shipped this in 1994 — as UFO: Enemy Unknown in Britain, as X-COM: UFO Defense in America, published by MicroProse either way — and thirty years of tactics games have been politely declining to reproduce it.
Two games, welded
The structure is two entirely different games sharing a save file.
Upstairs is the Geoscape: a rotating globe, real-time with pause and variable speed, where you build bases, staff them, launch interceptors at UFO contacts, and run a research and manufacturing operation funded by a council of nations that grades you monthly. Downstairs is the Battlescape: turn-based, grid-based, fully destructible, where eight to fourteen soldiers with names get off a Skyranger into a field at three in the morning.
The welding is what matters. Every Geoscape decision resolves as a Battlescape consequence and vice versa, with no abstraction layer in between to soften it. Research plasma weapons and your squad carries plasma weapons. Fail to intercept a UFO and it lands somewhere and does a terror mission on a city, and the terror mission’s score hits your monthly report, and a bad enough report makes a nation sign a secret pact with the aliens and withdraw its funding permanently. The global layer is not a menu wrapped around the tactics. It is a slow, compounding scoreboard that the tactics feed, and it can be lost quietly over four months of mediocre results while you are winning individual firefights.
That two-layer marriage is the design everyone copied. Firaxis rebuilt it in 2012 and built a very good game out of it. What Firaxis softened, deliberately and with a clear rationale, was the arithmetic underneath.
The tension is three systems interacting
X-COM’s dread is manufactured, and you can name the parts.
Time units. Every soldier has a pool per turn. Walking spends it, turning spends it, kneeling spends it, firing spends a percentage of the maximum, and crucially, unspent units are what fund reaction fire during the alien turn. So every action carries an invisible second price: what you spend moving, you cannot spend responding. The system makes caution and aggression into the same currency, which means there is no safe play — hanging back with full TUs means seeing nothing and letting the aliens choose the engagement.
Hidden information. The map is dark. Your soldiers illuminate a cone; the rest is black. Night missions are genuinely, structurally different from day missions because your effective sight radius collapses to about nine tiles and the aliens' does not. You throw electro-flares to buy visibility and the flares mark your position for everything that can see them. The information economy is honest and brutal: light is safety and light is a target.
Permadeath. Soldiers have generated names, they gain firing accuracy and bravery and strength by using them, and when one dies the stats die with him. There is no roster of clones. The squaddie who has survived eleven missions has become measurably better at his job through eleven missions of surviving, and that history is stored nowhere except in him.
Put the three together and you get the specific X-COM feeling that no later game has quite reproduced. A veteran with 70 accuracy opens a door on a dark farmhouse with four time units left. There is a Sectoid nine tiles away that he cannot see and that can see him. He dies during a turn you are not playing, to a shot you do not witness, because of a sum you got wrong ten seconds ago. The game never raised its voice. It just charged you for the door.
I have written before about what permadeath actually buys a designer, and X-COM is the load-bearing example: the deaths are not dramatic, which is exactly why they land. Nobody gets a scene.
The research tree is a horror mechanic
The bit people underrate is the research design, because it looks like a tech tree and behaves like an interrogation drama.
You cannot research what you have not recovered. Plasma weapons require plasma weapons picked up off a corpse. UFO power sources require a UFO you did not blow up on approach — and the interception minigame’s incentive is to shoot the thing down hard, which destroys the salvage, so the Geoscape and the Battlescape are arguing about your loot before the mission starts. And the deepest research branches require a live alien, which means building an Alien Containment facility, carrying stun rods or a small launcher into a firefight, and taking a Sectoid or an Ethereal alive on purpose.
Consider what that asks of the player. To learn the most important things in the game, you must voluntarily make a fight harder — swap a plasma rifle for a melee-range stun weapon — and then keep the thing you captured alive in a room under your base. The tech tree is gated behind escalating acts of nerve. Games have spent three decades looking for a way to make research feel dangerous, and Gollop solved it in 1994 by requiring a hostage.
The economy has holes, and they are worth admitting. Manufacturing sells at a profit — build laser cannons in a workshop full of engineers and you can fund the entire war without ever touching the council’s money, which trivialises the resource layer for anyone who notices. And the original DOS release shipped with a famous defect: pick any difficulty above Beginner and the game reset itself to Beginner after the first battle. A generation of players formed their opinion of X-COM’s legendary cruelty while playing it on easy. That is one of the great accidental facts in the medium, and it says something uncomfortable about how much of the dread was in the arithmetic and how much was in the dark.
The real ancestor
The lineage here is unusually clean, because it is one man’s.
Julian Gollop had been building this exact machine for a decade before MicroProse funded it. Rebelstar Raiders arrived in 1984, Rebelstar in 1986, and Laser Squad in 1988 on the Spectrum, the C64 and later the Amiga — turn-based squad tactics with time units, opportunity fire, destructible scenery and named troopers who stayed dead. I played Laser Squad on a C64 as a teenager without any sense that I was looking at a prototype. X-COM’s Battlescape is Laser Squad with a globe bolted on top and a research economy giving the squad a reason to exist. The Geoscape is the innovation; the tactical layer was already finished.
The descendants split into two families. Firaxis’s XCOM: Enemy Unknown in 2012 kept the two-layer structure and replaced time units with a two-action system, which is a real design position rather than a dumbing-down — two actions are legible, teachable and readable at a glance, and the 1994 TU pool is none of those things. The cost is that the door problem disappears. With two actions you can always move and shoot; the agonising fractional sum that made X-COM feel like a trap is gone by construction. The other family is the fan-maintained one: OpenXcom reimplemented the original engine wholesale, which is why a 1994 DOS game currently runs better than most things from 2014.
Where it stands
X-COM is a masterpiece with a broken economy, an unreadable interface and a difficulty bug that lied to everyone about what it was. It also generates more memorable events per hour than almost anything since, and it does it without writing a word of story. The narrative is emergent bookkeeping — a name, a stat line, a mission count, and then a gap in the roster.
Play it via OpenXcom, on a PC, at night, with the sound on. Then play Master of Orion 2 for the other half of what MicroProse was doing in the mid-nineties, or Civilization II for the same era’s argument about turn structure. And read save systems are ideology with X-COM in mind, because the ironman question this game raises has never been settled.
Spoilers below
The late game reveals what the whole structure was building toward, and it is psionics.
Around the point you have plasma weapons and personal armour and have started feeling competent, the Ethereals arrive, and they do not shoot at you. They take one of your soldiers. Mind control routes through a Psi Strength stat your troops possess and that you cannot see without building a Psi-Lab and testing them, and the test tells you, permanently and unhelpfully, which of your eleven-mission veterans is a liability. A controlled soldier turns and empties a plasma rifle into the squad he arrived with, and the game charges you for the ammunition.
This is the most vicious thing in it. Every other threat is external. Psionics makes your own accumulated investment — the veterans, the thing you have spent forty hours building — into the delivery mechanism for your defeat, and the only counter is to research the same power and use it on them.
The campaign ends at Cydonia, on Mars, in a two-stage mission that has to be won in one sitting with no reinforcement. The final chamber holds the Alien Brain, a lump of tissue in a room, and killing it ends the war instantly and everywhere. No cutscene worth the name, no farewell, no roll call of the dead. You get a short victory screen and a save file full of soldiers who did not make it to the end of a game that never once pretended their deaths were about anything.




