Collected Works
Elite’sl’ co-designer reflects on a career spent perfecting the galaxy
Elite ite co-designer David Braben reflects fl on a career spent perfecting the galaxy
ELITE
Developer David Braben, Ian Bell Publisher Acornsoft, Firebird Format (selected): Amiga, Amstrad CPC, Atari ST, BBC Micro, Commodore 64, MSX, NES, PC, ZX Spectrum Release 1984
ZARCH/VIRUS
Developer David Braben Publisher Acornsoft, Firebird Format Acorn Archimedes, Amiga, Atari ST, PC, ZX Spectrum Release 1987
FRONTIER: ELITE II
Developer David Braben Publisher GameTek Format Amiga, Amiga CD32, Atari ST, PC Release 1993
DOG’S LIFE
Developer Frontier Developments Publisher Sony Computer Entertainment Format PS2 Release 2003
ROLLERCOASTER TYCOON 3
Developer Frontier Developments Publisher Atari Format PC Release 2004
DISNEYLAND ADVENTURES
Developer Frontier Developments Publisher Microsoft Studios Format Xbox 360 Release 2011
RASPBERRY PI
Manufacturer Raspberry Pi Foundation Release 2012
ELITE DANGEROUS
Developer/publisher Frontier Developments Format PC, PS4, Xbox One Release 2014
PLANET COASTER
Developer/publisher Frontier Developments Format PC Release 2016
“THE CODE GOT SMALLER AND SMALLER; THE GALAXY DATA WAS ONLY ABOUT 100 BYTES”
Elite packed a galaxy into just 32K of memory, the BBC Micro’s heavy hardware constraints inspiring some pioneering invention from two self-taught Cambridge University undergraduates, Ian Bell and David Braben. After parting ways, Braben continued to apply his programming ability, fascination for science and technology, and keen commercial insight.
He founded what’s become one of the UK’s largest independent studios, Frontier Developments, makers of an eclectic range of games, including Kinect-powered playthings such as Kinectimals, detailed management games such as Rollercoaster Tycoon 3 and colourful platformers such as
LostWinds, while also embarking on projects like The Outsider, which were just too big for their time. He also helped to launch Raspberry Pi. Now, with 35 years in the industry behind him, Braben remains just as ambitious as when he started out, steering Frontier into a new era of selfpublishing which has restored the creative freedom which he enjoyed in the earliest days of his career.
ELITE Developer David Braben, Ian Bell Publisher Acornsoft, Firebird Format (selected): Amiga, Amstrad CPC, Atari ST, BBC Micro, Commodore 64, MSX, NES, PC, ZX Spectrum Release 1984
Elite was a product of limitations. Before I met Ian Bell I was playing around with the Acorn Atom. It was so hard to draw anything physical like a landscape, so space seemed very attractive because it’s mostly black and you can make it feel 3D by putting moving dots in. We called them dust because if they were stars they’d be travelling at stupidly high speed. One of the first programs I wrote was a 3D starfield in BASIC, and I was crestfallen by the fact that all these dots would appear and then disappear before the next frame was drawn. You needed a certain framerate to get a perception of movement. It was probably a good three months later before I figured out how to program in assembly language to speed it up. I did all the trade and planetary stuff in
Elite, as well as graphics, and Ian did combat and flying. His first pass at it was superb. Really good, beautifully coded and utterly, utterly impossible to beat as a player because the enemy ships used a tensor field around your ship to work out the optimal route to get behind and kill you, and you never saw them! We decided it was very good and also no fun at all. It was funny, we ended up with something much closer to the work I’d done before I met Ian, where I had it so a ship would come up, slow down and shoot you. We even had it coming down in front of you just because it looked good and you felt clever because you could get behind it.
But we were forever against the end of memory. We were both going through each other’s code thinking, “I can save two bytes there. Is it safe?” We did things that today would be considered utterly disgusting and unmaintainable, but the code shrank and we kept putting in new features. The procedural generation came about because we decided that we needed at least 50 locations to explore. So I asked Ian how much memory we had, and he said, “Well, none, but we can probably save 100 bytes, maybe 150.” So we had that to store the data for the stars, and I started to think, what if we used the first letter of the name to determine what type of place it is? Then, wait a second, if we generate them randomly… It was a gradual realisation. Suddenly, why only have 50, why not 250, or thousands?
The whole thing started flowing. We realised we could have an economy on each star because it didn’t need to be stored. The game stored where you were and just regenerated the details from scratch because it was so quick. If you think of it almost as an API to look up the market details, it went through the algorithm and would produce them, replacing all these data tables that we previously had. The code got smaller and smaller; the galaxy data that was held permanently was only about 100 bytes.
The code itself was extremely inelegant. People don’t face this these days, but we didn’t have enough space for the source code. It had to fit on one disk, because otherwise every time we did
a build we’d have to swap disks. To our shame, we knocked out all the vowels out of comments, then we took out comments if we felt the code was obvious. And then all of the variable names got smaller and smaller. So the code became utterly unreadable, but we were so used to it. Within a bigger company it would’ve been death, but we’d spent so long trying to make it smaller that we were so familiar with it that I still remember sections from it today.
When we took the game to the publisher, Thorn EMI, they rejected it. We got this letter which listed their problems which we thought were the key features of the game. They thought it was horrific that we let players save their position because the player would have to supply their own cassette. They thought it would stop people buying it, and they were very experienced, shiny suitwearing people, and we were just students. But when we showed it to Acornsoft a week or so later the response couldn’t have been more different, because they were gamers. They looked at it and went, “Wow, how have you done that?!” And, “Oh that’s brilliant, can I have a go, how does it feel?” Thorn EMI didn’t ask how it felt. Acornsoft were the people we wanted to work with.
The masters of Elite went off the week before my end of second-year finals, which was a hell of a dilemma. I was doing natural sciences, and Ian was doing maths. But what was really annoying is that was the year I did best in! But it was great fun. Our experience with Thorn EMI galvanised us into realising that there were lots of other platforms than BBC Micro. They wanted Amstrad, Spectrum, Commodore. That two- or three-year period was the Burgess Shales of computers, so many machines, and it was hard to know which were worth writing for. We decided to concentrate on 6502 CPUs first because it’s what we knew best, and did Apple II, Nintendo NES, Commodore 64 versions, and we subcontracted the Z80 versions out. It actually became a bit of a treadmill and we were working on a sequel as well, which never saw the light of day. Ian and I drifted apart a bit, which was a shame.
ZARCH/VIRUS Developer David Braben Publisher Acornsoft, Firebird Format Acorn Archimedes, Amiga, Atari ST, PC, ZX Spectrum Release 1987
When we worked on the sequel for Elite we were much more systematic, splitting up the work. But I got to the point where I needed a part of what Ian was doing and he hadn’t gotten around to it, so I suggested that I’d write something else for three months and come back. That’s what became Zarch, which later became Virus.
Acorn had given me a prototype of the new Archimedes. It was amazing to go from a hugely restricted 8bit computer to full-on 32bit with four megabytes of memory. I played around with the machine and saw what it could do, and I wrote a development environment for it. It compiled the whole of Elite in 0.1 of a second instead of 15 minutes, and it would also do ST, Amiga and Archimedes. I thought we could give it to other developers because Acorn were very worried that the machine wasn’t getting any traction. So I showed it to them and
said they could have it for free if they continued to support it. But they said no. “Why would we want to support our competition?” It was a bit annoying.
Anyway, it was a lovely machine. The CPU – I was still writing in assembler – was such a joy to write for. I had three months until the release of the machine to design, write and test the game and also to do Lander, a free demo. It was great! And it was very liberating because the thing built so quickly so I had a short iteration loop to make things feel good. But Zarch was a pig to play. The problem was that I got so good at it. I didn’t have time to go to artists, so it’s all programmer art and the landscape is generated by 10 sin curves at different directions so they all interfere to make the landscape, and below a certain level it draws water. It was really quick to do because I was constrained again, not by memory but by the time I had to do it. But there are a lot of things in that game I’m very proud of. It’s the first game to have directional lighting and shadows, proper algorithmic shadows, not drawn. It was the highest-rated game in ACE magazine.
Going back to do another 8bit version of Elite seemed pretty unappealing, especially as when I went back Ian still hadn’t done the other bits of the code.
Zarch, for me, was where I felt real freedom in solving problems. Afterwards we did several other versions of Elite; it ended up on 17 platforms. It’s a terrible thing to say, but when you do the same thing over and over you want to do something new. The last version of Elite that Ian and I did was the Nintendo one. It felt like it was the old world of 8bit and I wanted to go into the new world. Ian said he was interested in this and the other, and we said, well, why don’t we go our separate ways? That was when I kicked off doing Frontier: Elite II.
FRONTIER: ELITE II Developer David Braben Publisher GameTek Format Amiga, Amiga CD32, Atari ST, PC Release 1993
Frontier had a proper galaxy. It was so much better than Elite’s and I was really proud, and still am, of what is in there. It’s the real-sized galaxy, but it was a stick-thin sliver through the middle, so there were approximations and also there was remarkably little data to base the positions of the systems on. There was only data for 100 or so stars which were agreed by the various different data sources, and beyond that the data was so flakey that you couldn’t rely on it. I felt I’d be made to look an idiot when the accurate data came in, so I just left them out, and even stars I thought were quite accurate, they’ve changed. I have to confess that Achenar, for example, the capital of the Empire, is much further away in Elite Dangerous because we’ve got so much better data. To give contrast, there are 150,000 real stars in
Elite Dangerous, but even now studies are coming in with a whole light year difference in position!
I thoroughly enjoyed writing Frontier. People have asked why on Earth I went to all the trouble with the galaxy when it doesn’t really affect the game. And the answer is that it did affect the game because it kept me loving doing it. It was a sanity factor, because I found it pretty difficult to finish. Frontier took five years. What frustrated me was that I did all the engine stuff right at the start, in 1987 and 1988, and it was then stellar compared with everything that was out there. It could do all sorts of things, like rendering paragraphs of text on the side of a spaceship. That was in 1988, when almost all games were still bitmapped 2D games! But when it came out in 1993 there were other 3D games and I thought it was a shame. If I’d have shipped it earlier without all the galaxy stuff, it’d have stood out so much more.
“THERE WAS ONLY DATA FOR 100 OR SO STARS WHICH WERE AGREED BY THE VARIOUS DIFFERENT DATA SOURCES”
During the writing of Frontier I decided to form a company, and the reason we called it Frontier was to imply a carry-over and also because I thought it would be arrogant to call it Braben-something, because it feels tacky and that you could never step down, and for other people in the company it implicitly belittles them because their name isn’t Braben. Frontier stuck, and it was quite nice because one of the first products that Frontier made was a version of Frontier: Elite II for the CD32. That was a nice machine! It was probably one of the best versions of the game, but hardly anybody played it.
DOG’S LIFE Developer Frontier Developments Publisher Sony Computer Entertainment Format PS2 Release 2003
This was a really lovely game to have written, mainly because – and this is easy to say – but it was way before its time, which was a positive and a negative. I had always believed that we were appealing to such a tiny segment of people; thinking about my friends, would they play any of the games we’d worked on? They probably wouldn’t. We were going almost exclusively to blokes in a narrow age group, and I thought it’d be really nice to do something that could be enjoyed by a lot more people that didn’t involve killing.
I loved things that made you look at the world in a different way. I think Elite did that, and it was interesting to see how a dog perceives the world. I was thinking about what it would be like, when there was a smell it’d completely overload my dog’s vision. That’s where the black-andwhite and overlaid colours thing came from. We thought we could do multi-part puzzles where you can see footprints, all sorts of things like that, just to get people to think of the world as if they were dogs. I thought that was fun, and a whole new set of mechanics for puzzles became possible.
We started Dog’s Life in 1999, but it didn’t come out until 2003 because it was a bit start-stop. The problem was finding a publisher, which is why I say it was before its time. If you look at the whole game business, there’s an ebb and flow and things go out of fashion because there’s an unsuccessful game. We saw it with space games in 2003 with Freelancer, where you talked to a publisher and they’d go, “Oh, no no, space games are out of fashion now, they don’t do well.” But Sony thought Dog’s
Life could widen the market and give them something to talk about when they couldn’t talk about shooting games and so we went with them. They were forwardlooking, but they still said it was an awkward game to publish and I’m not sure the marketing team completely believed in it. It had a difficult birth but I’m proud of it. In a way we paved the way for
Nintendogs and I think it could’ve done a lot better if it’d had the visibility, or if we were self-published.
ROLLERCOASTER TYCOON 3 Developer Frontier Developments Publisher Atari Format PC Release 2004
Chris Sawyer was a friend and we’d worked together at a distance for a long time. He worked on Virus, Frontier: Elite II and Elite
Plus conversions, and he had this theme park management game called White
“I HAD ALWAYS BELIEVED WE WERE APPEALING TO SUCH A TINY SEGMENT OF PEOPLE”
Knuckle. Same as Dog’s Life, it had a very difficult birth and it appealed to a broad audience, which meant the marketers in game companies didn’t know what to do with it. “What do you mean it appeals to women? They’re not going to play games!” He ended up putting it through with Jacqui Lyons, our agent at the time. It was difficult going though Hasbro, who took it as a favour to Jacqui and thought it was a niche game. He got no advance or anything for it, and they made just a small number of units. It sold out immediately! They made more, and they also sold out. It was funny, I saw it again with The Sims, which EA did as a favour to Will Wright because they thought it wouldn’t sell. I remember a guy at EA joking about it, saying, “We’re doing this to keep Will Wright on board but it’s going to crash and burn.” Then only six months later he was saying, “Oh yes, I was thoroughly behind it.” We put a lot of effort into Rollercoaster
Tycoon 3 and it did very well and continues to, and had a 40 per cent female audience, even though we went 3D with it. At that time, graphics cards weren’t very settled and they were coming through in quick succession. I got criticism because we wanted to use pixel shaders. We got covers of most magazines because we had shiny reflective water and it looked so good, but people said, “But you realise you need a workstation worth $5,000 to get that?” I said that you just turn it to software rendering, and they complained that meant you didn’t get shiny water. Yup! But the point was that we did it to future-proof the game. I thought that the graphics cards next year would be half the cost. And they turned out to be even lower. We could see that rate of change, plus the fact that slightly cynically we knew we’d get better coverage if we supported the high end.
If you look at our industry over the last three or four decades, every year there’s been an amazing change for development. Analogue sticks on a console, the screen resolution going up, pixel shaders. All those things; when PC memory went up with a new Windows version; when we got DirectX. DirectX was a gift! Before that it was so hard to do a PC game. All of those things are big changes to our industry, and we’ve gotten into the habit that we’re not writing for today, we’re writing for next year. That’s what partly gave Rollercoaster
Tycoon 3 its longevity.
DISNEYLAND ADVENTURES Developer Frontier Developments Publisher Microsoft Studios Format Xbox 360 Release 2011
This was a very constrained development period of significantly less than a year, but it was working with Microsoft and we had a very good relationship with them by that point, following on from Kinectimals. The licence was very time-limited, so it was against the clock and we did a lot of outsource. We had nearly 390 people working on the game, and at that point Frontier was only 240 people. That’s a massive amount of outsource to manage, and Microsoft did it. I take my hat off to them for supporting us on that.
I’m very proud of the game in terms of the sheer amount of content, which has all of the different IPs from Pixar and Disney
“WE’VE GOTTEN INTO THE HABIT THAT WE’RE NOT WRITING FOR TODAY, WE’RE WRITING FOR NEXT YEAR”
that existed in the park, and it was delightful and eye-opening working with all the licensing and the famous actors for voiceover as well. There were some challenges working with Kinect but the game itself looked very nice. It was the whole of Disneyland, and don’t forget, it was on Xbox 360. Usually, we control draw distances to keep the framerate up, putting a hillside in here and there, but with Disneyland we had to have it spot-on. We worked to a very detailed survey, and you have to be able to render those actual sightlines. The one down Main Street was a challenge because you can see a lot of polygons for an Xbox 360, which was also running with the overhead of Kinect as well.
RASPBERRY PI Manufacturer Raspberry Pi Foundation Release 2012
In 2002 or 2003 we noticed the number of graduates applying to Frontier had fallen off a cliff. I talked to people at Cambridge and other universities and they said they’d had a big drop-off in applications to computer science courses, and it looked anecdotally like it was to do with the rise of ICT and the abandonment of computer science at schools. At the same time, we were doing focus tests with children of school age, where they’d volunteer during lunchtime to play a Wallace and Gromit game we were working on alongside Dog’s Life, and we’d ask them what their most boring subject was at school. Universally, it was ICT.
They hated it, and these were kids who volunteered to play computer games!
So I talked to people in government and tried lobbying to bring computer science into ICT, which was then about teaching PowerPoint, Excel and Word. But it was like pushing against a brick wall. I argued with Margaret Hodge [then minister for children] and she said I was special pleading for the game industry, which I didn’t understand. I was very annoyed, actually.
Later, I was at the founding meeting of a wonderful group called Computing At School where lots of like-minded people were looking at the problem. We looked at making a software platform, but when one brilliant teacher made an object-oriented programming environment, which allowed kids to learn how it worked, it wouldn’t compile so others could use it. This was a time when schools used different versions of Windows: NT, Me, 95. One guy said it worked on half the machines in the classroom but not on the other; you needed the right Service Pack, and this was when computer security started to come in and some virus checkers would stop it all working.
I spoke to a guy called Jack Lang, who became a co-founder of Raspberry Pi. He introduced me to more people, like Eben Upton from Broadcom, and the slightly bonkers suggestion came up that we make hardware. What stops us from doing that? Between us we felt we could do it. We then spent two wasted years trying to persuade the BBC to make it the new BBC Micro. We were going to call it the BBC Nano and there would be a TV show and we’d give the rights to the BBC, but they just weren’t interested. I wish they’d just said that. We had meeting after meeting, and in the end we had an exasperated call with Jack Lang in which we said, ‘Well, why can’t we fund it?’ It wasn’t going to cost that much.
So we put the money in ourselves and made 10,000 units. We thought we could sell it for $25, and we ended up having profit in that! Amazingly it came together, mostly down to Pete Lomas and Eben Upton. There were some pieces of luck in the hardware design which really helped. We used POP, package-on-package, where the RAM was in the CPU chip, which meant the circuit board was so much simpler, but then we realised it emitted way less electromagnetic radiation. That meant we didn’t need a case, making it much cheaper, and could still be compliant. That was a happy accident.
The first 10,000 were going to be developer machines so we could get a software base. My role was to publicise it; the BBC’s Rory Cellan-Jones had been supportive from the start and set up all these meetings. We did an off-the-cuff video that went viral, and then it became newsworthy because we got stupid preorders. On February 29, 2012, from five in the morning I was on all the BBC shows. I said we were going live at 6am, and a second later the server crashed and it had already taken all 10,000 orders. By the end of the day we were into the hundreds of thousands.
Raspberry Pi was designed to solve that education problem and also to make programming be seen as a positive thing. The dip we saw in applicants to computer science were now way overtaken; I think applications to computer science in Cambridge were up by a factor of six. We were without new programmers for more than 10 years and we put an end to it. The Department For Education brought computer science back to the National Curriculum and wound down ICT, and really helped, too. Computer science at school is interesting.
“I“IT TALKED TO THE GOVERNMENT GOVE ABOUT ABOUT COMPUTER C SCIENCE. SCIENCE. IT WAS LIKE LIKE PUSHING PUSH IN AGAINST AGAINST A BRICK BRICK WALL” WALL”
ELITE DANGEROUS Developer/publisher Frontier Developments Format PC, PS4, Xbox One Release 2014
The Elite Dangerous Kickstarter was a very interesting time because there hadn’t been an open world space game since Freelancer, which didn’t do very well. That’s why a lot of the big publishers had left the area clear. What was helpful was seeing just how big and dedicated Elite’s initial audience was. What was also funny was that if you looked at those people, like those in the queue to play at the EGX show we did after the end of the Kickstarter, their average age was in the 20s. They were too young even to have played Frontier when it came out! Obviously there was a nostalgia factor to Elite, especially early on, but it was quickly swamped by the general mass of gamers.
It was delightful to get back to Elite, to go and make the galaxy better and update it with the latest learning. We were also able to take a lot of scientific approaches I wasn’t able to before. We put in 150,000 stars, which is way more than you can see with the naked eye in the night sky. Roughly speaking, it’s what you can see with telescopes, but what was interesting was that we were populating the stars with a lost-mass model, because Hubble can only see roughly 40 light years and beyond that you can’t resolve individual stars. But based on the amount of light and the patterns, we populated those distances with roughly the right ratio of stars that comes from the mass model we have of the galaxy. Hopefully it’s pretty accurate. Amazingly, when the star TRAPPIST-1 was discovered, it was already roughly in the right place in our game and amazingly with the same seven planets! There was an amount of luck in the planets, but our simulation had plonked the star there. It shows that the simulation probably isn’t far out.
There’s also lots in the game that people haven’t found, like the aliens that were discovered in January last year. Part of you wants them to be found and part doesn’t, just because it’s fun to think there are big things out there that people haven’t seen. And there are other things that caught us by surprise, like the Guardian sites. We put in a lovely intricate breadcrumb trail that people could use to discover the location of alien ruins, and we did a sneaky video of in-game footage that showed an SRV coming over a ridge and seeing one of the sites. People got excited and what we hadn’t realised was that because it was proper footage the sky was correct and someone spotted Andromeda. Even though it was quite a long way out, they worked out roughly where it must be and then brute-force searched for a planet of its colour. It was joyous to watch, but it completely bypassed our breadcrumb trail! Now we’re very careful when we’re doing videos not to point the camera at the sky.
PLANET COASTER Developer x Publisher Frontier Developments Format PC Release 2016
The team that worked on Planet Coaster are most of the same team that worked on the original Rollercoaster Tycoon 3, which shows that a lot of people have stayed at Frontier for a long, long time. Planet Coaster was all about doing those things we knew could be done so differently and so much better with a clean sheet, and came with the challenge of making it look beautiful, giving players the chance to be creative and also making a great management game, and making sure that all felt like a coherent whole.
It’s been a very successful game, but it was funny. Because we’re a listed company, city people were sceptical that we could compete with the likes of EA and Activision. They said we had a free pass with Elite
Dangerous because of nostalgia, but could we do something new with something other than a brand name? They thought Planet
Coaster was going to crash and burn, and seeing us get lucky a second time helped address the city scepticism in Frontier.
We’ve added various expansions and put in lots of free stuff, so Planet Coaster has its own life and longevity ahead of it. If you look at the way the game industry has changed around sustained supporting of games, it’s enabled by the fact that we can go directly to customers and be very reactive. When people go all caps-lock on our forums, we can address the issue they’re unhappy about. As an industry, we’ve got almost a duty to make sure that people don’t feel miffed – that we build that longterm, loyal support.
“IT WAS DELIGHTFUL TO GET BACK TO ELITE, TO GO AND MAKE THE GALAXY BETTER AND UPDATE IT”