RAGE Productions

I was recently inspired by a reddit discussion to examine the complexities of branching narratives and their effects on game development. Consider a complex narrative such as Mass Effect, Elder Scrolls, or Fallout. These games all have one thing in common: choices the player makes as their character affect the world and create unique experiences based on prior decisions – or at least they appear to. The realtiy behind creating such a narrative structure is quite different from the end-user experience.

Using Scaffold 22 as a lightweight example, I’d like to examine how such a world and narrative is built, and how plot options and player agency are created.

Let’s assume a few basics: the world and setting are known and established, at least in theory. The basic narrative has been written. This was the state of Scaffold 22 after it’s first alpha release. There was a single-threaded plot. The game was linear. This is how we think about stories: as linear experiences – even if it’s a dynamic structure, we treat ‘our’ experience as a linear one. We don’t jump between options. We imagine a book, a movie, a memory. There’s a single progression of events that holds everything together. This is very much a manageable experience, both as a writer and a designer. You write in one direction, edit in one direction; everything lines up in the final version. That version can then be submitted to a publisher or editor, and they’ll be able to offer suggestions based on how they perceived the narrative.

Now let’s introduce complexity: a second, branched plot thread that completely deviates from the original premise. This never occurs in movies or books. It’s a phenomenon unique to interactive media, and one which requires a slightly different approach to handle. Submit a branching story to a publisher or editor, and they will be confused first and foremost. What is the story ‘really’ about? What’s the central theme here? Well, that can be hard to answer.

Technically, the issue is simple to examine. By introducing a branch, all those narrative elements we so painstakingly laid out in a linear fashion can (potentially) be invalidated by a player choosing to go an alternate route. In Scaffold 22, you’re presented with such a notable choice roughly 1/3 of the way into the game. Upon returning to or otherwise visiting the Riptide, a major club and party venue, you’re asked to choose between two factions:

This is a basic decision common to many branching narrative games. The question is simple: what faction do you identify with most? Whom do you trust? Which group do you want to associate with more? And what do you believe the consequences of said association will be? The player will make a choice and continue with the narrative altered to suit their personal tastes. A few flags get set in code to be checked later and development goes on. Sound simple, right? Let’s examine that simple binary choice in a bit more detail.

In Scaffold 22, the Raptor / Division 108 branch has a major effect on the world. Not only does the plot change by locking out or allowing different missions, but the characters you meet and enemies you might potentially face are altered. Different gear options are unlocked. Player progression and perception of the narrative is altered for the remainder of that play-through.

To achieve this sense of divergence, there are hundreds of dialog segments which have two versions: one for Raptors and one for Division 108. Numerous missions have two progressions: one for Raptors and one for Division 108. Narrative options and gameplay elements are made available or locked out. Simply put, the volume of total content in Scaffold 22 is, as a direct result of this choice, doubled.

There are five other faction branches and a few other major branches in Scaffold 22. Each of these has similar effects on the game, though none quite as noticeable as the Raptor / Division 108 split. Some are easier to find, others more secluded, but each and every major choice fundamentally alters the makeup of the game world and introduces exponential technical complexity to game design. New characters, altered dialog, new places, altered outcomes – major branches are, to the writer and designer, a major headache too. Consistency must be maintained and that means work. A lot of it. There is however a way to ensure this (over)load of content doesn’t become unmanageable…

As the name implies, a branch point is the point in time (and code) where a narrative branches between two or more options. Imagine a tree. Each branch leads to more sub-branches and so on. Pretty basic stuff so far.

In practice however, that tree is simply impossible to manage. There would be too much specific content resulting from each branch – a workload even the biggest development team couldn’t handle. We have to encourage the perception of such such a structure instead, and this is where branch points come into play. Any given branch point may or may not actually alter significant portions of the story – oftentimes, content is written entirely independent of the branch. But, in order to make the branch seem significant, it has to be presented well. A branch has to feel natural and not suspend disbelief. And the branch has to hide just how simplistic the narrative really is behind the scenes.

The most noticeable branch point in Scaffold 22 is the aforementioned Raptors / Division 108 split; it’s an overt change of the game’s entire tone. The game asks: do you want to maintain the more liberal status quo but accept criminality and corruption, or enforce a totalitarian, xenophobic rule that stamps out criminals and corruption by brute force? It also asks a few other questions and punishes some options but the point stands; this is a major shift in narrative tone, and a significant moment of character development.

Due to its significance, the event that spawns this choice is located at a major plot hub: Wilkin’s Riptide, the club I mentioned earlier. There is no way the player can avoid this branch. It’s simply not possible – well, except for in that one very specific plot progression which no one will find without a game guide. By branching so obviously, the player is forced to deal with this choice, and that is tied into the tone of the branch: this is a MAJOR and OVERT decision which EVERYONE in the world will feel.

What about those other faction branches? Well, they’re a bit more subtle. Three of those are criminal subfactions who prefer to stay out of the limelight. As a result, their branch points are hidden in the narrative. And the fourth faction is your original employer – there isn’t even a noticeable branch there; simply following the main plot line without doing anything to change the status quo will ‘branch’ at the appropriate point, without the player ever noticing.

See how location of branch points are significant to the tone of the narrative? Status quo means nothing is changed. Criminal factions in the shadows are hidden and hard to find. Overt actors are presented as such to the player directly. By placing these branches at the right places, the player is building their perception of the game as a series of diverging choices that result in a tree-like structure.

In actuality, they’re narrowing down their narrative to a specific tone by choosing a specific path (imagine: chopping branches off the tree). If done well, and the narrative flows properly, they’ll never realize they’re actually doing this – or that the tree is simply a (well placed) sequence of choices that may or may not actually relate to one another. We’ll get to all that in more detail later. First, we have to take a moment to examine how these branches are presented to the player…

Obviously, branching willy-nilly and at arbitrary points in time doesn’t work. The player will be confused and not care about the choices. So the narrative has to lead up to every potential branch – or at least try. You can never quite guarantee a player will care about a branch, or even realize what’s going on.

The aforementioned Raptor / Division 108 split? Well, depending on progression that can fly right past the player. They’ll be surprised by the event. That’s actually intentional, as the narrative builds up the choice as a surprise raid by Division 108, timed so that it (hopefully) occurs at a point where the player is used to the status quo. It’s shock and awe factor is supposed to jar the player out of their complacency and force them to consider the wider implications of their choices on the world. This is in fact a trick: there aren’t more than a dozen. But, by presenting it as such, we can lead the player to believe their future choices will be SIGNIFICANT! They’re really significant in lower-case letters and with a whispered consequence, but don’t tell anyone.

In other cases however, branches are introduced more subtly, often with a quest line leading up to the choice, allowing the player to ‘get to know’ the options they’re dealing with. Depending on the tone of the choice to follow, it’s introduced as a mystery, or an investigation, or a designated sub-antagonist. That’s pretty much standard narrative structure and good writing practice. Not much to say here from a story-telling standpoint.

But there’s another issue: every major branch risks locking out content. In the above example, let’s assume we chose Wilkin’s Raptors. Suddenly, all Division 108 related content is unavailable. One third of the way into the game, 50% of the content has been locked out by a single decision! The game’s narrative couldn’t care less, of course – the player made their choice and is now dealing with the consequence. From a gameplay and development perspective however, that’s bad. Not only do we have to write more content, but we have to carry double the load for every choice. Multiplied by dozens of other choices and it’s impossible to handle. Time for another trick.

In order to avoid locking out too much content, we present the character with branches – and an option to opt out or into the consequences. Even the MAJOR! branch in Scaffold 22 isn’t quite definitive. You can slaughter Division 108 and still meet their employer. Or you can wipe out the Raptors and still pursue the main plot line without their meddlesome influence. There are side quests and modifiers to the main quest line designed to handle both cases (actually, a variation on the same mission that is re-used in several other places) and only one definitive point per branch where the player has to decide if they do or don’t want to go that route.

Gameplay and narrative wise, this is achieved by separating the branch point from the point of absolutely no return; that moment when the player decides to make a change to their personal narrative from which they cannot recover and cannot back out of. Every faction has one of these, and every major decision can be opted out of (or into) at various points. Opt out of every option consequently and the narrative defaults back to its ‘normal’, linear progression as originally written and eventually ‘locks in’ to that storyline.

Lock in somewhere else, and the other progressions become unobtainable. This allows the player a degree of agency despite the fact they’re already narrowing down their options. They can make choices, suffer consequences, and still decide how they want to handle a situation. This isn’t so much relevant to the narrative (as all excluded or not yet triggered branches are simply ignored) but very important to gameplay and the player’s perception of progression. Without this phased ‘locking in’ of various choices, the player would be constricted too quickly, and feel like they’re being railroaded down one path without any real agency.

Of course this is still all a trick. They’ve already opted in or out of major choices. But, by giving the player a little leeway, we’re allowing them to breathe, to decide, consider, re-think, and ultimately settle for a given progression, not when the game forces them to but when they want to.

We’ve talked a lot about the basics in abstract terms, and how they cause expoential workload for both the writer and developer, but let’s look at some specific techniques and tricks which can be used to lessen the workload without harming the experience.

Somewhere in the depths of Scaffold 22 is a character named Commisaria Varai. She appears in several events related to criminal behavior, a single side mission, and the lead-up to the ‘default’ ending. She can be killed. She can come to tolerate you. Or she can become a tertiary antagonist (seriously, she’s an annoying character).

Given the number of potential outcomes of interacting with her, we want to avoid tying her into the main story line too often. Why? Because, if she’s dead, that means someone else has to take her place. If she tolerates you, that requires unique dialog, and if she hates you, another set of lines for that eventuality. Multiply those cases over the total number of times Commisaria Varai appears in the narrative, and consider there are several other characters with similar reactions, and one quickly realizes why her narrative must be isolated to prevent content bloat.

As a result, there are only three places where Varai appears: the criminality events, that one side mission, and the lead up to the ending. She’s significant from a narrative perspective (especially if you meet her and come to hate her) but she’s deliberately excluded. There are a dozen scenes where her intervention would be interesting from a narrative perspective. But including her would cause a massive content creation overhead. So her narrative is isolated. There is only one intersection with the main plot line.

I use this technique a lot in Scaffold 22 to avoid choices spilling over into impossibly convoluted situations, where hundreds of conditions need to be checked to build a simple scene. Sub-plots are most often isolated entirely from the main story, barring occasional reference. This significantly reduces the workload and keeps the narrative manageable.

I hinted at this earlier, but virtually every choice in Scaffold 22 is binary opt-in or opt-out. This sounds simplistic, but bear with me. By making sure the vast majority of outcomes have only two states, I reduce the number of eventualities I have to cover as a writer and designer.

To hide this simplistic structure, numerous binary choices are layered on top of one another. Let’s return to the old Raptor / Division 108 example for a moment. A player who chooses the Raptors and then chooses to speak to Division 108’s employer will have a vastly different experience from someone who chooses Division 108 and kills their employer.

Viewed on their own, each of these choices is binary. Layered together, they form a complex pattern and help solidify the ‘tree’ of choices the player seems to be making. There’s no need to define each branch of the tree in detail however. Instead, we design situations in a ‘void’ and contextualize them slightly with a few throwaway lines of narrative prose.

Playing Scaffold 22, you’ll rarely notice this is happening. Pick the code apart and you’ll quickly realize that you aren’t really being given as much freedom as you’re led to believe. This is what I meant earlier with restricting options; narrowing down the narrative while leading the player to believe they have more options than they do. That’s a core element of writing interactive stories and branching narratives. The player’s experience has to be narrowed down so that we (as humans) can perceive a single plot thread – our experience. The trick is doing this without telegraphing the reality of the situation to the player and making them feel like the experience is hollow or forced.

Another useful trick related to exclusion is inconsistent plot lines. In Scaffold 22, if you go with Division 108 and pursue their plot, you immediately invalidate any and all content which relates to the player pursuing the default (Church of Eden) narrative. And vice versa. Viewed in a void, these two narratives are wildly inconsistent: pursuing the Church ending, you’re liked by character X. Go another route and X will despise you. Objectively these stories cannot coexist. The events don’t add up and, if you were to change a few variables at runtime, the narrative would fall apart.

Since that doesn’t happen in practice however, we can simply treat those two plots as entirely different progressions of the underlying narrative. By being mutually inconsistent, and by ensuring those two plots don’t have to coexist, we can reduce the number of contextualized content that needs to be created.

Opt out of a Church of Eden ending, and you’ll never find a loving, fatherly character X. They simply don’t exist in that version of events. Go the Church route and that character will magically re-appear in the narrative at the appropriate point in time. This technique is used extensively in Scaffold 22, with numerous consequences simply being thrown under the bridge and ignored, to avoid having to deal with them.

Provided there are enough other progressions to hold the player’s interest, this shortcut won’t become apparent, as the player is still facing the cumulative consequences of all their actions. That character X happens to be ignored entirely up until the point they become relevant again… well, the player’s imagination will fill the gap.

I have a general rule about encounters: there should be at least three options available to the player at all times. Scaffold 22 goes out of its way (most of the time) to implement this. By offering more than a binary choice (even if the result is binary, see above) the player is forced to think a little more about how they want to act.

Even a simple yes/no choice can be enhanced with a third option: tell me more. The player will still be forced to decide eventually, but the third option allows a little more freedom. They can opt to decide immediately or learn a bit more before committing. These sort of options (even if they are bogus) are essential to maintaining the illusion of agency – and contextualizing choices, especially when there really isn’t anything to choose between.

Unless a specific event has to be explicitly mentioned, don’t mention it. This avoids having to (re)write countless scenes for every possible eventuality. Any given character or event that is not the focal point of a discussion can be referred to vaguely as ‘that guy’ or ‘that thing’, allowing the player to infer significance based on their personal experience. This cuts down on workload again, while maintaining narrative consistency.

I use this technique a lot in throwaway dialog and random referrals to previous events. Who cares if the player chose Raptors or Division 108 when we’re talking about their emotional issues? Simply mention “past events” and let the player decide what those events really were based on the story they experienced. They’ll probably not have forgotten. And that’s one condition check less that needs to be implemented every time the topic comes up.

This is also used with discussions of backstory. A player who hasn’t read Scaffold 22’s extensive Codex entries will be able to infer what might have happened, while one who has will immediately associate mentioned events with what they read. In the case of backstory and fluff, this technique is in fact a necessity – I can’t guarantee the player picked up on the significance of any given fluff article even if they read the Codex entry (which I can’t reliably determine anyways), so let’s just be vague about it and hope for the best.

Another way to avoid overloading the story with intersecting branches is offer content that is of purely mechanical value, but has little to no narrative significance. A good example of this in Scaffold 22 is the ‘Highborn Palace’, a slaver den that the player can rule over or exterminate – or simply leave alone. There is only one plot-relevant narrative branch that occurs here, and one tangentially important quest option. Beyond that, this entire location is divorced from the main narrative, and exists within its own little microcosm that’s occasionally affected by external events but never in a significant way.

Gameplay wise however, Highborn Palace offers access to a new player class, new gear, and an alternate source of income (and a boss fight to test one’s gear and skills). By divorcing this location entirely from the rest of the story, we don’t have to ever include it, and never have to check how the player handled discovering the location (if they even did). There are two other locations in the game that follow this pattern, and numerous ‘flavor’ encounters that do the same.

They feel signficant, and are mechanically and to the player experience, but don’t actually affect the narrative in any way. This allows content to be added, offering places to explore and characters to meet (or kill), without complicating the narrative structure. You could call this ‘padding’ but it isn’t that simple. By creating these micro-narratives, without referencing the main story, the game world can be fleshed out and the experience enriched, and narrative bloat is conveniently sidestepped.

Numerous options presented to the player – options that are significant to the player’s perceived experience – are purely mechanical. The weapons they carry. The outfit they chose. The augments they installed. None of these are checked by the game’s narrative except in very specific and highly situational circumstances. By offering these customization options to the player however, the player can define who they are as a character without affecting the narrative – or the narrative affecting them.

You can play through any story line in a fine business suit. Or you can grab a minigun and wear the heaviest armor in the game. Neither will affect the plot noticeably. But it will affect how the player perceives their experience, because the game’s mechanics behave differently based on gear, skills, etc. This immediately deepens the narrative experience and sense of character development. The story of Scaffold 22 would work just as well with a cardboard cutout. It wouldn’t be a mechanically fun game, but the story would be consistent. Properly integrating mechanics, experience, and narrative without any one stepping on the others toes too much is essential to ensuring the player isn’t shoe-horned into a given role – unless they chose that role deliberately.

Creating a branching narrative and integrating it into satisfying gameplay is a challenge for any developer to say the least. It requires a lot of patience, experimentation, shortcuts, tricks, and a good sense of how complex systems interact – not just on a technical level. Branching narrative is as much about emotion and player experience as it is flags set in code.

In-game events have to be rich, varied, unique. Gameplay has to be consistent. Narrative has to be narrowed down to a single, coherent plot. Everything seems to be pulling in different directions and yet the gameplay experience as seen by the player has to be consistent. Viewed from a distance, all this may seem insurmountable and impossible to handle. Broken down, with a few dirty tricks to help along the way, it’s entirely doable even by a small team.

Scaffold 22 isn’t the odd one out. It’s the result of one developer who wanted to figure out how to write a branching, narrative-focused role playing game. Sure, it may have taken a little longer than some games, and may not be the most technically impressive product, but I consider a two-year development cycle to be reasonable for the product it produced. I had to learn a lot along the way though, iterate over content many times, until I found a system that worked.

I hope that this article will shed a little light on the mystery of branching narratives, and help fellow developers create similarly complex works (be they text-based such as my game or fully interactive, 3d worlds).