The Signal From Rossignol: Some Thoughts On The Development Of The Signal From Tölva

Jim Rossignol, project lead.

In this post I’m going to talk a bit about how we made The Signal From Tölva. There are few points I want to make along the way, but one of the most crucial is about how this game, like so many indie games, represents the creative instincts and ideas of the people who made it.

In the years since the launch (and remarkable success) of Sir, You Are Being Hunted (2013) I’ve occasionally been asked for lessons, thoughts and opinions on game design and development. I’ve been a little reticent about coming forward those things and that’s been because I don’t really feel like there are (m)any universal lessons to learn from our experience.

We’ve made esoteric first-person games without worrying too much about what other games did (although our inspirations and aspirations are obvious enough) and the bootstrapping steps to success that led us here seem unrepeatable: getting a break with a chance contract from Channel 4, catching a wave with Kickstarter’s novelty and excitement in 2012 – a progressively trickier task as the years have progressed – releasing with the first wave of Early Access games on Steam, arriving with the first wave of survival games (our announcement pre-dated DayZ by just a couple of months). All these things seem to amount to singularity in spacetime: impossible to draw a generalised rules from.

All of that caveating done, I should say that I nevertheless feel compelled to talk about our development experience as it reflects on our latest release, The Signal From Tölva, and so I shall do so.

This is that game.

The Signal From Tölva is an open-world shooter that in some ways resembles landmarks in the genre (it certainly shares genetic material with STALKER games and Far Cry games), but it does so in a way that takes advantage of our peculiar strengths and interests as a development team. It’s an AI-driven experience, in which the systemic activities of the AI deliver the action, with little in the way of scripted encounters. This has both advantages and weaknesses, as I’ll discuss, but I think the most important aspect is that it has a specific feel and a distinct personality. Things that I find essential for any game in 2017.

The personality of Tölva is, as astute people have observed, the personality of the people who constructed it. It is odd. It is remarkably tranquil, even contemplative, particularly for a game that also leans heavily on combat elements. This surprises many, but it is by design. It is a pure expression of the interest I personally have in the escapist terrain that video games can provide. The original working title for Tölva was “Highlands”, and the initial design sketches took ideas from hiking in British and non-British wildernesses, particularly those which featured bleak, treeless canyonscapes. A visit to Iceland fell quite early in the development cycle. Quite obviously, games – like other creative projects – are direct functions of the abilities and interests of the people who made them. But this somehow seems super-true of The Signal From Tolva which feels like an amalgam of what I like, how James thinks, what Olly can art, and how Tom codes. (And also how Dan fixes everything we shonk together.) It is an alloy, with the new properties that such a material exhibits.

To some extent larger projects are able to hire for specific roles and skills in a way that is beyond the limited budget and reach of a small team like ours. In the case of a multi-million dollar development: if there’s a job that needs doing, then someone appropriate is hired to do it, and their work is managed so they hit (or at least approach) a specific target. The results of this are clear in some of the best AAA games: they’re able to repeatedly hit proven templates, even as teams shift and change. A huge aggregate of creative work is made, where some individuals will stand out, but where ultimately the team effort is what is delivered.

Not so with the tiny gang of omni-devs. The reality here is somewhat shiftier, with small teams often trying to hit a moving target that is moving only because of the nature of the process they’re engaged in. The template doesn’t exist. The plan changes as the game comes into being. People spark off each other and the things that get made. That process is one that can be incredibly fertile and outrageously difficult. One of the things I am personally most proud of is what the small Big Robot team has achieved, given its minimal size and its lofty targets. I believe that came about because of who we are, as much as, say, which tools we used, or which middleware we chose to mutilate to our purposes. Importantly, this isn’t the part where I go so macho and talk about 100-hour weeks, because while we’ve worked hard, and done some serious time, we’ve not worked so hard that we caused ourselves grief. We’ve made sure burnout wasn’t a thing. (Which it has been for us in the past. We are old men now, and have learned our limits.) We’ve been extremely ambitious, without allowing those ambitions to crash the project. Some of the principles that we applied to make that happen are discussed in this Gamasutra interview. In short: we made things modular, we made tools that worked, we relied on live systems and emergence over scripting and engineering. We made a game that knew its limits. Tom will talk more about this in other posts.

This brings me to some creative specifics that I think are important to the project, because they were important to individuals. This angle on things helps me understand what we made, and so perhaps it will help you, too.

1- The Complicated Legacy Of The Game Before

The Signal From Tölva is the second album (AVSEQ and Fallen City being rare singles/EP releases, in that metaphor). It is the response to Sir, You Are Being Hunted, and the experiences we had during the making of that game. Where Sir was procedural, this was built by hand. The productivity of a procedural system has been replaced by the control of a man-made one. Sir was super lo-fi, and Tölva has been produced to a far higher standard, with art assets that wow. But also there’s something about the world creation that has stayed similar despite the radically different processes at work.

Tom and I were responsible for the decision to make the world ourselves, and that decision did seem counter to Tom’s wide-ranging interest in generative approaches. Although Olly and myself did most of the world editing, it was Tom’s response to finishing the Sir project that led to us taking this route. You might have expected him to vote for a procedurally-generated approach this time around too, but he felt differently when it came to commencing development. He erred on the side of control, arguing that we should try to make a game that was both prettier and less variable than anything we’d done before. Tom’s interest in game design is dominated by an extraordinary inclination to tackle all things himself, but also by a pragmatism and anxiety about actually getting it made. Not having to do procedural work this time gave him more time to work on literally everything else, and that basic fact made the target of ‘prettier’ seem far more achievable within the time frame.

Tom’s approach to letting me construct a 4km x 4km world was to set things up in a way that could be easily be comprehended. In fact, I was facing much the same task as the algorithm in Sir, You Are Being Hunted, because Tom created real-time tools that echoed the automated structure of that world. Our game level was once again divided up into a cellular grid (perceptible on the map screen, but not so much in the world itself) in which I could assign props to specific cells, so that they might be loaded systemically, and lodded in and out appropriately depending on your range from the location. This of course made the task seem more manageable, too, because I could tackle one cell at a time, making sure they blended into each other organically and naturalistically as I worked.

The tool I had to do this was real-time, too. Unlike normal Unity scenes, where you have to run it to check your edits, we did all our editing live, by having nothing more than a terrain heightmap and skybox in the scene itself, and then saving all the props to an XML file. This meant I could run around the world in one window, while placing and saving the locations of props in another. It also meant that we could use the XML to save and load all kinds of data about the world, without having to save that into the scene itself. Where a scene of this size would become impossibly unwieldy in Unity’s editor, our system made editing both a playable experience, and relatively performant (it does crawl on slower machines now the full world is implemented). This approach, I think, is emblematic of Tom’s background: he is an artist who taught himself to code. He might lack some of the engineer’s sensibilities when it comes to structural work, but it was instinctive for him to treat the level designer like he treated the automatic level designer in Sir. At the same time provide a system that allowed hands-on experience of the world I was creating.

As I know Tom plans to discuss elsewhere, his approach was similarly practical when facing the task of getting the rest of this huge game done. Making things systemic and to some degree “build themselves” was a key aspect of the solutions we came up with, even without an automated generative process building the world itself. The robots know how they are equipped and behave accordingly. It’s a modular system where the AI uses the same equipment as the player. The weapons in their hands are literally the same as the weapon in your hands, because you are one of them. Fiction coheres with technical and man-hour limitations. It works like this because how else do you make a system that creates enough variety to populate a game world when you have so few resources for systems, art, and design? Modularity. Scaleability. Having the world populate and run itself. These are all key shared interests, and approaches that come naturally to the BR team.

Critically, though, it was what was best in Sir that we carried with us to Tölva. We knew that we loved the dynamic nature of AI entities clashing in an open world, and so everything we did was designed to facilitate that in this new game. Tölva’s three factions are spawned in squads which have specific goals in mind: to guard a location, to attack a location, to explore, to survey a wreck or a ruin, or to attack a location held by antagonists. They react to events in the world, investigating sounds and engaging enemies. They encounter each other and the player in a fluid, freeform way, and the experiences the game produces reflect that. This also means that a single playthrough is never enough to see what the game can throw up: a strength and weakness. What if a player could have had a more finely tuned experience? But what if they experience a dynamic and emergent situation which was only true for them? What treasure! Heady, important stuff.

For development, too, this has ramifications. I’ve played this map a thousand times, and thanks to the way it throws stuff into the mix, I am still facing surprises.

2- The Philosophical Value Of A Clear Blue Sky

We got lucky with the unlikely singularities, as I mentioned, but they happened during Tölva development, too. Fortune (and a dash of boldness) meant that we got to work with concepts from a master, former Rockstar art pro Ian McQue. Just as significantly, we hired a 3D guy (this was Olly’s first game!) who actually had the wizardry in him to make the most of those materials and to create a coherent art style for the game world. The results speak for themselves, but here’s a video of Olly speaking about them anyway:

Here too, our personalities dominate the final outcomes. My personal reference point in first-person games might be STALKER, but I never felt like we needed to do the grimdark skies of the genre. Why not create an unsettling and glitchy weirdness under blue skies? Why not create a sense of unease punctuated with ultra-violence in a place that is just nice to be? Why not rely on the eerie and the weird, rather than the grim and bloody? Why not just clear skies, when everyone else’s are filled with clouds?

The blue skies of Tolva are also emblematic of our desire to create a game world that captures a specific feeling and an identifiable look. Olly regularly cites games like Journey and Abzu, with their unmistakable palettes. Our skies are perhaps not entirely faithful to the Ian McQue concepts that we otherwise worked so hard to emulate, either. Instead the look is very much an example of how a visual style coheres around a bunch of elements, both technological and artistic, and exhibits personal responses both to other games, and to the process of developing this one. We went through numerous iterations, featuring everything from psychedelic cloudscapes to a vast, shattered planet in the sky, eventually returning to the clear blue gradient which fades to a starfield as the night progresses – a result of using a dynamic skybox to facilitate a clean night/day cycle.

Ultimately I feel like this single decision – not a cloud in the sky – symbolises the entire Tölva development process: one in which we worked to create something that was both iconic and beautiful, but which is ultimately esoteric within the expectations of the genre. A deliberate iconoclasm is evident in the personality of everyone who works at BR (how the hell else would we get projects like this started?) and I think this symbolises that most vividly.

3- Hacking (Gun) Existing Systems

As development went on, interaction with and utility of the numerous allied NPC bots was the vast robot elephant in the landscape. It had to be mentioned, but we did not speak of it. I had some nebulous ideas about how to achieve the interactions we knew we’d need to make the game work, but I continued to focus on other things. James (Carey, design lead on Sir, You Are Being Hunted), always true to form, tackled the problem headlong, building the command module (a ‘weapon’ that allows you to recruit and command allied bots) in a matter of days. Utilising the existing Ai systems to implement the concept, he constructed a system whereby bots can be removed from their existing units (and therefore existing AI missions and priorities) and forced to follow you or investigate locations that you direct them to. James has a long-standing interest in squad-based tactics and Ai-player co-op interactions (he even worked on the Arma 2 single-player campaign, which is a landmark in such gaming) and so it seems inevitable that his outstanding contribution to the game should be this system. Although devised fairly late in development, it has heavily influenced design since then, while at the same time being an optional tool within the overall game. It is entirely possible to get through the entire experience without ever equipping this entertaining tool, but of course you’d be missing out (or possibly challenging yourself) if you do so. (Again this seems symbolic of our attitude towards the player: most games would have forced the tool into your hand. As with fiction and exposition, we avoided that. It’s your choice, go explore for yourself. The rewards are always greater that way.)

The Command Module implied all kinds of additional details: the unique IDs of the bots allowing you lament their individual loss*, for example. It also gave us a wider range of tactics for the player to engage in without us having to do much more in terms of building new systems. We only had to modify existing AI to make it work. Everything was keyed off the NPCs that we’d already built to bring the world to life, and it relies solely on the abilities of the bots to navigate the world and engage enemies. The Command Module provides another good example of our need to make do with limited resources, and to build within the small garden of systems that we had already planted.

(*Losing those bots, however a small or routine and event, certainly created the response that James was looking for, which was not simply a slightly broader set of interactions, but an emotional reaction to working alongside AI entities, however speechless and fleeting their existence might have been.)


The Conclusion, Such As It Is

The most exciting thing for me about The Signal From Tölva is also the most difficult thing: that it is not necessarily what it seems to be, and that the expectations people bring to it aren’t necessarily matched by the reality. That’s a communications nightmare, of course, and it’s been tricky to really get across what people should expect. Games are so often about clear message and here a clear message was hard to do. This was absolutely a tension within development, too, with different influences and desires pulling the game in a number of directions. Nevertheless I feel like the final game landed a clean blow: it stands tall within the genre of open-world shooters, perhaps because of (rather than despite) being made by a five-man team. We punched above our weight and feel proud of that. And, regardless of its small size, the game offers a bright chunk of escapism that I am still returning to even after hundreds of hours of play. The Signal From Tölva is a strange, quiet game in a landscape of extraordinary multi-million dollar noise. Under the familiar interface of “a map full of icons you can travel to” resides a weirdness that I personally would like to pursue further, and a sense of beginning to explore a frontier that I find enormously alluring.

To take a slightly more philosophical stance: games are often very literal, when they could and perhaps should be more figurative and metaphorical. I’ve found it hugely interesting and exciting to see people discuss the meditated or “prosthetic” nature of the Tölva experience: that you are not the robot you are controlling, but simply an entity remotely linked to it. There is no huge depth or insight in such a trick, but it fixes simple things, like fast travel, and it makes people break step and think for a second about the world they’re engaging with. In this and other things The Signal From Tölva has an ambiguity to it that I think is singular, and which I personally savour each time I play it. I can’t wait to delve further into the fictions we produced for the game and I am pleased that the lorebook is just the start of that.

Next, though, we continue to build. There are things to fix and other formats to look at. And then there’s an expansion on the horizon, too.

Hmm. How ever will we approach that?