Sir, You Are Being Hunted, and the advantages of procedural generation

Big Robot's first-gentleperson stealth game Sir, You Are Being Hunted is remarkable because it represents an indie developer's attempt to explore and master the possibilities of open-world procedural generation and AI, two aspects of game design that seem sorely under-appreciated in contemporary big-budget development. Moreover, Big Robot made a point of developing it in public, using Unity and funds collected from a very successful Kickstarter campaign plus the money earned from making the game available on Steam's Early Access service last August. Sir officially launched this month, and the fact that Big Robot pulled it off suggests that an independent developer with an interesting idea for a project can still make it happen without relying on a publisher -- as long as they're willing to commit to keeping a conversation going with the fans who support them. Big Robot founder Jim Rossignol expounded upon the studio's trials and triumphs building Sir and exhorted his fellow developers to continue delving into the possibilities of procedural generation during an email interview with Gamasutra. Here are the highlights.

What inspired the design of Sir?

I suppose that most games are a vast mishmash of possible influences, because they can eat up so much media. The thematic thrust of Sir is in the British landscape and in old sci-fi, but there are so many other influences -- from my interest in spooky ambient music, to programmer Tom Bett's interest in generative mathematics -- that I doubt I could list it all. James Carey, our designer, regularly quotes the influence of the 1940s movie Kind Hearts and Coronets, which he'd watched just before we embarked on work on Sir. You can see that sort of darkly comic feel throughout the game. Britishness mocking itself, all that good stuff. Loads of things like that got folded in, because it's all just fuel for the development engine. The real inspiration, though, was our interest in non-linear and systemic design in games.

We're not really interested in making a tightly engineered and strictly plotted experience, although we love seeing people make those, and instead we wanted something with as much freedom and systemic chaos as possible. Although Sir has a beginning, a quest (of sorts) and an end, we were really just interested in seeing if we could make a totally open world with AI roaming around doing their thing, giving the player a bunch of different tools and abilities, and then letting those systems spark off each other to create interesting situations. Nothing is scripted in Sir, and even the level designs aren't fixed, so everything has to emerge systemically. That's the sort of game that fascinates all of us, and it was the challenge of how to build that as a first-person game, and how to make it serve a dark, funny concept like humans being hunted by robots, that really inspired us to tackle it. It was enormously tricky to produce and balance something so esoteric, so we're rather pleased with the end result.

This is the first game that Big Robot has funded via Kickstarter and published on Early Access. How did that affect the development process, and what would you tell another developer embarking on a similar project?

Kickstarter and Early Access were both astonishingly useful for us, both raising money and putting enough eyes on the game for us to get the feedback we needed to create a stable, functional game. That said, I am not sure we can offer ourselves as a representative example of how those platforms work, or should work, not least because I think they -- and the attitude of gamers towards them -- change all the time. The main issue, I think, is the extent to which you are required to commit to both these ways of raising funds. You can't just lay down your terms and expect to raise money, and you can't just go quiet and work once you've got the money. You have to put much more time in, on videos, in forums, on blog posts, and on generally running this living process that you've sparked. This is a huge, dynamic, ongoing conversation where you have to illustrate what you are doing, explain what you are doing, and expect feedback and opinions on what you are doing. Both Kickstarter and Early Access are about doing what you are doing in public, and you have to really embrace what that means to get the most out of it. Anything less means you are going down the wrong path.

Any particularly amusing or infuriating problems you ran into while making Sir?

The most consistent trial, and ultimately a key triumph, was in the development of the AI. The robots of Sir are formidable, but only because they can see, hear, and hunt successfully. Getting all those elements functional didn't actually take that long, but then refining them for the right "feel" and smoothing out all the bugs took most of the development time. There were some deeply obscure issues that emerged, such as one refactoring of the animation systems fixing a (hitherto invisible) bug so that robots were then better able to maintain line of sight. I think they had previously been looking away from the player when reloading, but now they never lost you -- and since they always ran to the last place they saw you and searched for you, the hunt became brutal. With the vision bug fixed, the robots went from dangerous to unstoppable, and the design team then had to rebalance their aggression and vision variables to return to the "feel" we'd previously settled on with the bugged system. Another near-invisible issue came with the robots having a different stance in certain states than others. When patrolling they were upright and marching about, but when hunting their head (and consequently their eyes) went forward. They would therefore occasionally (and imperceptibly) clip through scenery when in motion, and this had the effect of the robots glimpsing the player through solid objects, but only for a moment. So we had this irregular blip of robots screaming "WHAT IS THAT?" even though they should have had no way to detect the player at all, and immediately lost sight of their target. Once sorted, though, we believe Sir's robots make some of the best adversaries in any stealth game. They don't just give up as soon as you dive into cover, and their ability to hunt and search for you right across the island is exactly what we'd been aiming for. They create a sense of vulnerability for the player, while still being a bit slapstick.

I'm fascinated by procedural generation in games; tell me a bit about how the procedural generation systems in Sir work and what decisions influenced their design.

The reason we pursued procedural approaches within Sir was that our main programmer, Tom Betts, has always been interested in this sort of approach. Before we made Sir he was building procedural terrain systems (including one infinite landscape engine I'd still like to return to!) and during the development of Sir he was still experimenting, delivering a prototype game for free after last Christmas (Permutation Racer). So it seemed natural to us to look for solutions in the kind of area we were interested in. We knew we wanted to make an open world game with a lot of terrain, so generating that seemed like the logical direction to head in. What's interesting about Sir's approach is the constraint we gave it: that it be recognizably British landscape. This means fields, woods, crumbly old villages, collapsing farms and so forth. It ended up being a fairly broad caricature, but the consequence was that it made something which was immediately recognizable. Creating that was a fascinating mix of technical and artistic challenge, especially as we had no full time artist! As for the question of whether it was a worthwhile time investment compared to having

someone build the levels by hand, it depends on your point of view. There have certainly been people who have played Sir and then asked what difference it made, and why we didn't just build one level, but personally I now find it weird that other games do stay the same each time I play them. That's not to say that Sir's procedural generation was about replayability, because I think that's just a neat side effect, but that the generation of the terrain as a thing in itself is a fascinating process and product for me. Turning off the AI and just wandering the landscapes has given us (and numerous fans) a joyous second aspect for the game. (Yes, we can be a Walking Simulator, too!) That many people don't even realize that the terrain is generated really says something for the power of Tom's system. Regardless, though, we had always intended to build an automated level designer, and so that's what we did. I don't think it's going to put anyone out of work, but there was never really a question of us as a team creating this world by hand, because that's not where our interests or expertise lead us. The success of the system we made that means we can now create 5 km of consistent, playable game space in about a minute. The new 1.1 patch that we're working on right now puts a lot more of that stuff in the hands of the player too, so radically and thematically mashed levels are now possible. If nothing else, that's a fascinating toy, and a successful experiment on our part.

Going forward, procedural generation systems seem likely to grow in popularity and power. How do you think that might affect the industry at large?

I think like much of what is going on in game development at the moment, it comes down to tools. Procedural generation systems, for whatever sort of game element they are being applied to, are simply tools fashioned to a specific end. If we've learned anything from the trajectory of engineering games it's that tools get refined for particular purposes, and then eventually become very powerful and easy to use. A small village created by Big Robot's British Countryside Generator. Tools such as those that create terrain procedurally are already getting enormously potent and accessible, and hopefully that will begin to reach into other areas, such as procedural texturing or procedural animation. What needs to keep happening, though, is that people need to keep pushing the boundaries of what they need to use procedural systems for, and how to apply them in their games. People don't build refined tools for systems they've never seen before, so it's down to developers in all walks to keep imagining how procedural solutions might be applied to game systems, and then making the prototypes that can inspire the tools that will become available to everyone in years to come. Hopefully we'll do a bit of that as we make more games.

How do you and your Big Robot compatriots feel in the wake of Sir's release?

Relieved. The game has done well enough for us to continue working on the particular esoteric flavors of games we enjoy making, even though the way it was funded - via Kickstarter and Early Access - meant that most of the money we were earning with it was immediately spent on development. The great risk was that we'd not make enough overall to be able to be able to fund another game, but that hasn't happened. We're looking forward to talking about what that means for us in a few months time.