Midwinter: The Survival Sim Before the Genre

Every survival game since has taught me the same lesson in the same order: the map is the enemy, the meters are the story, and the moment you feel competent is the moment the design has finished with you. Midwinter got there first, in 1989, on machines that could draw perhaps a few hundred flat-shaded polygons a second and had no vocabulary for what it was doing. There was no genre to put it in. Mike Singleton built it anyway.

Singleton had form. The Lords of Midnight on the Spectrum in 1984 had already pulled off the trick of making a small machine feel enormous, by drawing landscapes from a library of stamped-down elements and letting the player’s imagination handle the joins. Midwinter, developed by Maelstrom Games and published by Rainbird, took the same ambition and pointed it at a 3D engine. The Atari ST version came first; the Amiga and PC followed. What arrived was an island the size of a grievance, buried under snow, and a player character with a body that got tired.

The island is the antagonist

The premise is a light dusting of science fiction over a very cold problem. A meteor strike has thrown the climate into a new ice age. The island of Midwinter is one of the few habitable places left, its Free Villages organised into something like a functioning society, and General Masters is invading from the north with a mechanised army. You are the Peace Force. There are fifty of you scattered across the map, thirty-two of them playable, and none of them are in the right place.

What matters is the geography, because the geography does the design work. Midwinter’s terrain is filled polygons — ridges, valleys, plateaus, the occasional building — rendered in a palette of white on white with a grey horizon to tell you which way is up. It is austere to the point of abstraction and it is exactly right. Snow flattens detail in real life too. The engine’s poverty and the fiction’s premise land on the same image, which is the sort of accident you only get when a designer builds the world and the renderer to fit each other.

Crossing it takes time. Real time, measured in game hours, and the game keeps a clock that Masters is also using. His forces advance whether you are doing anything useful or not. So the map stops being scenery and becomes a budget: the ridge between you and the next village is a decision about how many hours you can afford, and the decision has a cost in a currency the game tracks on your body.

Fatigue, temperature, injury: the meters that meant something

Here is where Midwinter earns the label nobody had for it yet.

Each character carries fatigue, body temperature and injury as live values. Ski downhill and you go fast and tire slowly. Ski uphill and you burn yourself out. Take a hang-glider off a ridge and you cover ground for free while your temperature drops, because you are stationary in freezing air with no exertion to warm you. Sleep restores fatigue and costs hours you may not have. Injury slows everything and heals on its own schedule, which is to say slowly, and being shot is a logistics problem that lasts for days.

Those systems interlock in a way that modern survival design usually fumbles. The common failure is meters that only ever punish: hunger and thirst as a tax on existing, decrementing while you do the interesting things, refilled by a chore. Midwinter’s meters are levers you pull. Body temperature is the price of the fastest transport in the game, and you pay it deliberately, knowing what it buys. Fatigue is the reason you take the pass instead of the peak. The state of your body is the reason you pick one character over another for a given errand, because the geologist who is already exhausted and forty miles from the fight is a different resource from the fresh one sitting in a village nobody has threatened yet.

That is the design craft worth stealing. A meter earns its place when the player chooses to spend it. Everything else is bookkeeping.

Thirty-two people and a persuasion problem

The recruitment system is the part I still think about, because I have not seen anyone do it this way since.

You start as one character. The other thirty-one exist on the map as people, each with a location, a set of skills and — this is the move — a web of relationships with the others. Friendships, marriages, family, old grudges. To recruit someone you have to reach them and persuade them, and your chance of success runs through who you are and who they already trust. Send the wrong person and the answer is no. Send someone their sister likes, having already recruited the sister, and the door opens.

The consequence is that the game has a social graph as its progression system, and the graph is a puzzle laid over the geography. Recruitment becomes an order-of-operations problem across an island where every leg of the journey costs hours and body heat: reach the right person early and three others come cheap, reach the wrong one first and you are stranded with a list of people who have no reason to listen to you.

Each recruit is then individually playable. You can switch to any of them at any time, which makes it a strategy layer with a first-person sim running underneath — a shape the phrase “action adventure” was nowhere near covering in 1989. Send one character to mine a bridge while another skis a warning to a village while a third sleeps off an injury. The clock runs for all of them at once. Lose one and the graph loses an edge, and the people who trusted them get harder to reach.

The nearest modern cousin is the ensemble in an XCOM run, where the value of a soldier is the hours you have poured into them. Midwinter goes further, because its characters had value to each other before you arrived.

What it cost, and what it bought

I want to be honest about the friction, because Midwinter is an uncomfortable thing to play and the case for it survives that. It is slow. The interface is a wall of icons that assume you have the manual open, which in 1989 you did. The polygon engine runs at a frame rate best described as deliberate, and hang-gliding across a white void for ten minutes to reach a village is a genuine test of whether you are enjoying the concept or the game. The combat is thin — a vehicle sim with weapons, sufficient rather than good.

But the friction is doing something. The reason the island feels enormous is that crossing it is a commitment, and a fast-travel button would liquidate the entire design. The reason the meters matter is that you cannot bulldoze them with abundance. Midwinter is a game about being outnumbered on foot in bad weather, and every irritant in it is downstream of that being true.

Compare it with Hunter two years later, which took a similar polygon island and pointed it at sandbox comedy: steal a bike, steal a boat, cause chaos. Both games are open worlds by any modern definition, arriving before the term was in use. Hunter chose freedom as fun. Midwinter chose freedom as pressure, and pressure ages better.

The ancestor it shares with everything else on this shelf is Elite — the same insight that a machine with no memory can hold a huge world if it stores rules rather than places. Midwinter applies it to terrain and to people: a heightfield and a relationship table, and the game is what happens when you drag a tired body across them. The Sentinel was doing the parallel trick with filled polygons and dread on the C64 two years earlier, which tells you the technique was in the water and the application was the hard part.

Why nobody followed

Midwinter II: Flames of Freedom arrived in 1991, bigger and less interesting, which is the usual arithmetic. And then the idea went quiet for roughly fifteen years. Survival as a genre came back through mod culture and early access, through zombie mods and crafting trees, and it arrived with hunger bars and inventory grids and almost none of Singleton’s actual argument.

The argument was this: a survival system exists to make travel mean something. Midwinter’s temperature gauge is there so that the shortest route and the safest route come apart, and you have to know which one you are taking and why. Strip that out and the meters degrade into chores.

The other half of the argument — that the people you recruit should have their own opinions of each other, held before you showed up — is still sitting there unclaimed. Thirty-six years on, a game with thirty-two playable characters, a social graph you have to solve, and a body that tells you when you have overreached is still a description of something nobody has quite built again.

Play it on an ST or Amiga emulator, with the manual open, and give it an hour before you judge the frame rate. The island does the rest.