Frontier: Elite II — The Universe That Fit on a Floppy

Frontier came on a floppy disk, and inside the floppy disk there was a galaxy with roughly a hundred billion star systems in it. That sentence has been doing the rounds for thirty-odd years and it is still the most efficient advertisement any game has ever had for its own central idea. It is also, once you have spent a fortnight inside the thing, a decent description of what went wrong.

David Braben spent the best part of a decade on the sequel to Elite, published it through GameTek in 1993, and the Amiga and Atari ST versions came first, with the PC following. The Amiga had been my main box for six years by then, and it was exactly the right hardware to receive this. The pitch was that everything the original had faked with elegance, Frontier would now do properly. That turned out to be true. It turned out to be the problem.

What “properly” meant

Start with the flight model, because it is the thing everyone remembers and the thing everyone argues about.

Elite’s ships flew like aeroplanes. Push the stick, the nose goes where you point it, the ship follows, and if you cut the engines you stop. None of that is how anything works in a vacuum, and every space game of the era did it anyway, because it is legible and it makes dogfights feel like dogfights.

Frontier throws that out and gives you Newton. Thrust in a direction and you accumulate velocity in that direction. Rotate and you are now pointing somewhere else while still moving the way you were. Kill your velocity by turning around and burning against it. Orbits are real: get your speed and altitude right and you will circle a planet indefinitely, get them wrong and you will describe a beautiful arc into the surface. There is a set-speed autopilot assist for people who want out, and a manual mode for people who want in, and the manual mode is a genuine spaceflight sim in a game you bought in a cardboard box in 1993.

It is exhilarating for about a week. Then you notice the arithmetic. Real orbital mechanics across real interplanetary distances means a Hohmann-style transfer takes days or weeks, so Frontier gives you time acceleration — you can wind the clock up by enormous factors, watch your ship coast, and drop back to normal time when something happens. Which means the honest answer to “how do you fly in Frontier” is that you set up a burn, fast- forward through the consequences, and correct at the far end.

That is a real thing that real spaceflight does. It is also a game where the skill expression is largely in the planning, and the middle is a fast-forward button. The simulation is impeccable and it has quietly eaten the minute-to- minute play.

The stars are the actual stars

Here is the detail that still makes me sit up, because it is the purest statement of what Braben was after.

The galaxy is procedural — generated from rules, the same trick the original pulled to fit eight galaxies and 2,048 planets into a machine with 32K, now scaled up until the number stops meaning anything. But the region around Sol is different. Nearby star systems in Frontier are drawn from real stellar catalogue data: Sirius, Barnard’s Star, Alpha Centauri, Wolf 359, Tau Ceti, sitting at their real positions and real distances, with the fictional Federation and Empire built on top. Fly out from Earth and the first hundred or so systems you meet are ones an astronomer could check.

That is a design decision with an argument inside it. The procedural galaxy is a claim about scale, and any large number would have served. The catalogue is a claim about truth — that this is our sky, and the reason the Barnard’s Star run feels different from a random generated hop is that you know the star is up there tonight. Braben spent his budget on making the near neighbourhood verifiable, and the whole galaxy inherits the credibility.

Then you land on the planets. Actually land — descend through the atmosphere, pick a spot, put the ship down on the surface of a world that was generated seconds ago from a seed. Elite had space stations and nothing else. Frontier gives you gravity wells, atmospheric entry, a landing pad or a patch of rock, and the terrain underneath is fractal and continuous and mostly, in 1993, grey. Twenty-three years before No Man’s Sky did the same thing with a marketing department, this was on a single disk.

The thing the simulation ate

So the technology is a triumph, and I want to be exact about what it cost.

Elite’s genius, on reflection, was compression of a specific kind: it compressed a fantasy. You were a trader with a laser and a cargo hold, and every system you visited was an economy with prices, an authority with an opinion of you, and pirates with an opinion of your cargo. The 32K forced Braben and Bell to keep only the parts that generated decisions. Everything in Elite was there because it made you choose.

Frontier keeps the fidelity and loses the density. The trading model is deeper — a real economy, ships with mass and fuel and a delta-v budget, military and civilian markets, a rank system for two competing polities. And the moments per hour collapse. Getting from a station to a hyperspace point is minutes of fast-forward. Combat, under Newtonian rules with no arcade handrail, is two objects on ballistic paths trying to shoot each other while their relative velocity does something unhelpful, and the honest verdict is that it is more interesting to describe than to fly. Missions are text. The galaxy is vast and it is vast in the way a spreadsheet is vast.

The design principle underneath is one I keep meeting on this desk. Fidelity and engagement are separate axes, and past a certain point they trade against each other. Every increment of realism you add to a simulation adds seconds of dead time to the loop, and dead time is where a game leaks out. The Newtonian model is more truthful than Elite’s aeroplane physics in every respect, and it makes the ship harder to feel, because feeling a ship requires a tight relationship between an input and a response.

You can watch other designers hit the same wall and swerve. Midwinter simulated fatigue and body temperature four years earlier and kept them fast, because Singleton made each meter into a choice the player spends rather than a condition they endure. Hunter built a polygon island out of the same family of technology and then filled it with things to do every thirty seconds. Frontier had the deepest engine of the three and the emptiest minute.

Why it still matters

None of this is a dismissal, because Frontier’s failure mode has turned out to be more useful than most successes.

Look at what came after. Kerbal Space Program is Frontier’s flight model with the trading and the empire stripped out and the failure made funny — proof that Newtonian flight is a whole game if you build the game around it rather than bolting it under one. Elite Dangerous, Braben’s own return in 2014, walked a chunk of the physics back on purpose: flight assist on by default, a speed limiter, a combat model that flies like an aeroplane again, with the full Newtonian mode available for people who ask. That is a man who spent twenty years thinking about it and concluded the arcade handrail was load- bearing. No Man’s Sky shipped a procedural galaxy with seamless landings in 2016 and spent the following years adding the density Frontier never had.

And the compression claim stands up completely. The galaxy fits on the disk because the disk holds the instructions, and the instructions are smaller than the thing they describe by a factor with no useful name. That is the insight the original bought with its 32K and Frontier proved had no ceiling. Every open world since has been drawing on that account, usually while shipping a hundred gigabytes.

The reason to load it up now — and it emulates perfectly well on an Amiga or a DOS box, with community ports floating about for anyone who wants it native — is to feel a design being pulled apart by its own ambition in real time. The best ten minutes are the first landing: atmosphere, terrain resolving under you, thrusters fighting a gravity well the game calculated rather than staged. The next ten minutes are a fast-forward. Both of those are the same decision, made by the same person, for the same reason.

Braben got what he asked for. It turned out that a universe you can verify and a universe you want to live in are built from different materials.