Thinking about game balance as part of narrative design (and why you should care)

When we published a large balance patch for our game this October, it was met by a lot of surprise from our player-base. For those of you not in the know, Scarlet Hollow is a choice and dialogue-driven visual novel, and in the words of one player on Twitter “they did a balance patch for a visual novel lol.”

thank you UwU

Why do balance patches matter for your story?

Normally, when developers talk about a balance patches, they’re talking about altering the mechanics of their game — usually by tweaking some of the math on the back-end— so some playstyles become more effective and others less effective, balancing them out. There are a few reasons they might do this:

• Their game is multiplayer-focused and they want to make it more fair.
• Some playstyles being more powerful than others creates FOMO for players interested in less powerful playstyles.
• Players naturally gravitate towards whatever is “strongest” or “best” in a game, meaning an unbalanced game might lead to content not being experienced. For example, a class being more powerful than its peers in an RPG might result in weaker classes being picked less, and by extension, development time spent on those weaker classes less valuable (and the experience less rewarding for the players).

Scarlet Hollow, of course, does not have multiplayer or combat or any mechanics beyond picking dialogue. There’s some added nuance with how the dialogue trees function, but fundamentally, you see a list of options and you pick one.

Uh oh.

But we also have a “trait” system. At the beginning of the game, you get to pick two modifiers that define your character — you can choose to be observant, or physically fit, maybe Hot if you’re feeling spicy, and so on. These traits offer additional scenes, give you an occasional “get out of jail free” card for difficult decisions, and offer unique insights to the mystery underpinning the game.

For the most part, traits manifest themselves in our decision trees. But they also sometimes passively change how the game works, either by altering your relationship with other characters behind the scenes, by changing narration or NPC dialogue, or by automatically shunting you into an exclusive scene.

One of the less subtle passive changes introduced by the “hot” trait.

Let’s talk a bit about dialogue trees — when you let a player pick dialogue options, you’re doing a few things:

1. You’re defining their agency. Any decision in any game is defined, on some level, by the competing forces of you telling your player what to do and you letting them decide what happens. How you strike the balance between those two things over the course of a game will ultimately change how your player interacts with your story and will often influence how much they care about what happens.
2. You’re asking them to make a decision about who they are, as a character.
3. You’re asking them to minmax their world state.

Much like the first point is in a constant state of conflict with itself, the last two points are in a constant state of conflict with each other. Games more often than not teach players to pick whatever option seems “best.” This is great for decisions that mostly impact mechanics (do I make a second barracks or train three more marines in Starcraft), but it’s often anathema to good storytelling when it comes to non-mechanical decisions.

In these cases, when a major story decision arrives, players often wind up asking themselves “which of these options creates a better world state” rather than “which of these options is what my character would pick” or “which of these options makes for a more interesting story.” This is doubly bad when your story offers unequivocally “good” or “moral” options that come without consequences — these options are often not only weak from a storytelling perspective, but encourage players to think about the story in a way that removes themselves from the narrative, a sort of player-character ego death, if you will.

To use an example of an almost perfect moral dilemma and narrative decision that winds up missing the mark, I’d like to talk about the major decision in Redcliffe in Dragon Age: Origins. Beware, ye reader, for here there be spoilers for a twelve year old game.

The Arl of Redcliffe is unresponsive and ill, and the town is under siege by undead and other monsters. Upon investigation, it turns out that his son, Connor, allowed himself to be possessed by a demon to try and save his father’s life, and the demon is now wreaking havoc, controlling the castle’s court and attacking the town. Upon discovering this, you’re confronted with two options:

1. Kill the child (the game wants you to know that this is immoral)
2. Use blood magic to sacrifice his mother (a willing volunteer) to confront and exorcise the demon (supposedly dangerous, and probably also immoral.)

It’s an interesting dilemma with no “good” solution, but then you get option three:

3. Go send for a mage from somewhere else who will let you exorcise the demon without any consequences.

Kill the child. Kill the Mother. Or kill neither and suffer no consequences? Hmmmmm.

Based on how powerful the demon possessing Connor has shown itself to be, it seems like option 3 should have consequences, given that it would take time to travel to find a mage, and time again to return, during which the demon might break free from its restraints. This doesn’t happen, though, making the third option in this scenario unequivocally moral and better, and training players to assume that there will be no consequences to “doing the right thing.”

But if there are no consequences to “doing the right thing” and there are consequences to making “immoral” decisions, the significance of player choice is diminished, and your game’s narrative instead becomes a morality play. This is a similar issue in other Bioware titles, like Mass Effect, where paragon options have no long term repercussions, despite the fact that Renegade decisions are supposedly defined around being morally questionable yet pragmatic.

Another issue unique to these decision-making mechanics is that a balancing factor is often the perceived physical risk a decision puts you in. But physical risk means nothing in a medium where the only consequence of failure is that you have to try again until you succeed. Given the disconnect gameplay mechanics often have with narrative, this isn’t a risk at all, nor is it ever perceived as one.

Which brings us back to one of the main questions of this article: why should you focus on balancing player decisions in a choice-driven narrative game? Why is that something that needs to be balanced at all? I think it comes down to a few things:

1. Without balance — without felt consequences, without tangible pros and cons — the illusion of choice evaporates, and your decisions are so meaningless that they might as well not exist.
2. Clear right and wrong decisions encourage players to engage with your narrative mechanics as though they’re gameplay mechanics, breaking immersion and lessening the emotional impact of your story.
3. Content takes time to make — if you’re making content for a route a tiny fraction of players will ever actually see, you’re wasting development time. The way that additional content obfuscates the illusion of choice is vital, however, and it’s better to make branching content that players will actually see instead of not having it at all.

For an example of a game which tackles this remarkably well, see The Witcher 2, which has a dramatically different third act based around what decision you make in Act 1, and where major decisions have no right answer, so players are more evenly distributed across content.

Game balance philosophy in Scarlet Hollow

Now that we’ve laid some groundwork, it’s time to talk about how all of this ties into Scarlet Hollow. We have a few core principles we do our best to stick to:

1. Every major decision must have no clear-cut right answer and there should be clear consequences on the other side.
2. Each branch of a major decision should be something a non-trivial percentage of our player-base takes. Strive to split the room, and accept something that at least 20% of people wind up taking if there are only two options. We’ve had varying success on this front.
3. If an overwhelming percentage of players all gravitate towards one option, they should at least feel conflicted about their decision.
4. For lighter decisions, make characters feel like people and obfuscate back-end data to minimize min-maxing. Aka, hide the relationship matrix so Stella feels like a person you’re taking to instead of a set of numbers you have to maximize.
5. Redirect player attention to how their character is feeling to minimize metagaming and maximize roleplaying.
6. Traits should have relatively even pickrates, even if it’s not perfect.

Starting with the big stuff, every episode has at least one major decision between two bad options. As a means of giving traits extra oomph, and as a way of further highlighting the consequences of those decisions, each trait has (or will have) a specific situation where it can be used as a get-out-of-jail-free card. This still leaves five situations like this in a given run where a player is forced to pick between two bad options, not counting possible other situations where no traits convey an advantage.

We’ve had mixed results in player splits on that front.

The first big one

The Episode 1 decision fell under our margins: of our players who didn’t have Powerful Build, the trait that lets you save both Duke and Gretchen, a little over 16% saved Duke, and a little under 84% saved the dog. This split was originally within our margins, but introducing Talk to Animals as a trait led to 15.4% of our players switching from saving Duke to saving Gretchen. So without that trait, we’d instead be looking at a 31/69 split, which is closer to what we’re looking for. Given the popularity of Talk to Animals and how it’s wound up impacting the rest of our game, its influence on this decision is a trade-off we’re happy to make.

Worth noting that the flashlight and dog option is reasonably close to the Powerful Build pickrate in our most recent official survey, so this is likely an accurate sample.

Good thing nobody’s the sort of bastard who picked Talk to Animals and THEN decided to kill Gretchen

The major decision in Episode 2 was a lot more lopsided, with 89% of players opting to go deeper into the mine to try and save Alexis and Becka, and only 11% cutting their losses to get Rosalina.

Granted, this decision was the one that granted Keen Eye (our most popular trait) a third option, so the sample of players who actually had to make a choice was only about half of our player base. Still, that margin was too severe for our tastes, so parts of that decision were changed as part of our balance patch.

Love to crash here after a long night of almost dying in the woods.

Major social decisions have had much better splits. Episode 1 ends with you choosing to either return to the crumbling Estate to stay with your sour-faced cousin, or stay in your new best friend’s guest room, which has a 54/46 split in favor of your cousin.

Episode 2 has three major social decisions — first, whether you travel with Stella or Kaneeka to the mines, which has a 34/66 split in favor of Stella. This is a decent split, as she’s the character players have spent the most time with, and it’s comfortably within our preferred margin.

The latter two are how you approach investigating your cousin’s mining camp, then whether or not you bring her with you for the dangerous events of the evening. Those are both three-way splits, and we’re pretty comfortable with how they played out.

If you’re interested in a deeper dive on those decisions, please check out our earlier post on them!

The last chunk of decisions worth poring over is the traits themselves. There are seven in total, and players are made to pick two of them in a given run. In an ideal world, this means we’d see a 28.5% pick rate for each trait, assuming they’re balanced in such a way where all of them feel powerful to use and all of them appeal to similar sized subset of our player-base.

Our Episode 2 survey results were… not that.

Powerful Build, Street Smart, Book Smart, and Hot were all under-performing. This was particularly troubling when it came to Powerful Build, given that its major power spike comes in the first episode, so, if anything, players should be biased towards it.

Part of the skew came from the mega-popularity of both Keen Eye and Talk to Animals. I talk about why we think that is at length in an earlier post, though the bottom line is that, Talk to Animals is fun, does what it says on the label, and feels like it would have the most exclusive content gated behind it (it probably does.) And Keen Eye is a combination of the most obviously useful trait for a mystery game as well as the option that is least character defining, making it an easy second trait to pick.

Still, on top of that survey feedback, there were a few things we noticed more anecdotally about those four traits — streamers would often comment about the options not lining up with their expectations, and that, at least in the case of Powerful Build, Street Smart, and Hot, we were intentionally holding them back out of fear of making them too powerful. Our rationale was this:

1. Scarlet Hollow is a horror game, and letting Powerful Build make you feel too physically capable would weaken the horror elements for those players.
2. Scarlet Hollow is a mystery game, and letting Street Smart make you too good at catching people in lies and getting into places you shouldn’t be could ruin the mystery elements, both for those players and others who might hear spoilers
3. Scarlet Hollow is a romance game, and letting hot be too good at that would make the romance elements too complicated. We also wanted some of the effects of hot to be less noticeable to players (like being hot in real life, being hot in Scarlet Hollow gives you huge modifiers to your relationship with other characters, often without you realizing it.)

Balancing act

On the trait side of things, the most important thing was to become self-aware of the self-imposed limitations we outlined above. Once we realized we were holding ourselves (and our players) back, it was relatively easy to provide creative solutions that heightened the appeal of those traits without breaking the game. First, we asked ourselves, for each of these traits, what’s the fantasy that pushes someone to pick them?

1. Someone picks Powerful Build because they want to be strong. So we need to lean into that and create moments where that strength can thrive. And we can integrate those moments into the horror rather than thinking the two are opposed to one another.
2. Someone picks Hot because they want to be hot, and because they’re probably more invested in the romance elements of Scarlet Hollow than other players. So we can lean into that and give them the romance they crave, rather than hiding how much of a benefit Hot is to them. If Stella is instantly sweet on you, Hot players should know it.
3. Someone picks Street Smart because they want to feel witty and underestimated, and because they want to pick apart the secrets of Scarlet Hollow’s mysteries in a different way than Keen Eye players want to. If we can provide opportunities for Keen Eye to get extra information, we can do the same for Street Smart.

So, bit by bit, here are some of the major changes we added for each trait:

Powerful Build

The save Powerful Build gives you in the first episode is more… athletic than strong, which is fine, but it doesn’t capture the power fantasy of the trait. So on top of being able to heroically grab Gretchen before she runs off with one arm and a flashlight in the other, Powerful Build gets an opportunity to clock a Ditchling in the face, and to brag about it later.

squelch

We also added options for Powerful Build to more successfully fight Wayne in Episode 2, gave non-Powerful Build players an option to try and fight him and get trounced and added more opportunities for them to explore by lifting things.

Street Smart

We gave Street Smart more opportunities to sneak around without getting caught, including the ability to complete a stealth section in the second episode without getting caught while still getting as much information as possible, something other traits are incapable of doing.

We also designed a section in Episode 1 that lets them explore some of the sealed off areas of the Scarlet Estate — originally, players ran into a locked door that required a key to get in, which felt terrible for Street Smart players, so we’d patched it so they were able to open that door only to be met by a brick wall.

Look… we were pressed for time XD

Considering that the description for Street Smart literally says no door can hold you and the first locked door you come across mocks you with a brick wall, we knew we had to fix this. So we gave them access to a couple of extra rooms with unique early information that other players don’t have access to. And to stop them from exploring the entire house, we presented them with insurmountable physical obstacles that felt like they actually belonged in the wing, rather than were thrown in by lazy devs (like the brick wall.)

That’s more like it!

Hot

Initially, we set a methodical, slow pace to romance in Scarlet Hollow that would be more accurately reflective of knowing your love interests for less than a week. But if we were sticking to those guns across the board, what even is the point of playing a Hot character? We added a lot of extra flirting, and made confident flirting options exclusive to hot characters. They also have options to move relationships forward a lot more quickly — they can sleep in the same bed as Stella at the end of Episode 1, for instance (no funny business, though! We gotta keep ’em wanting more.) They can be a lot more direct with Oscar and Kaneeka as well. The further we get into future episodes, the more pronounced these changes will be.

Just lying in bed together and quietly staring into each other’s eyes. Like friends!

We’re also planning to accelerate the romance components in general — originally, our plan was for all romance lock-ins to happen at the same point in time in Episode 6, but we’re going to start including optional early lock-ins for players (hot and unhot alike) starting as soon as Episode 4.

Combos

We also added a few more specific interactions for trait pairings that we felt went particularly well together. Hot and Powerful Build gives players some “himbo” style options (regardless of gender), and Street Smart and Powerful Build give players some powerful options as well, both in terms of having a stronger skill set while sneaking around, and being more clever about how they maneuver in a fight. Most trait pairings have at least one combined option, though some have lent themselves more towards these combos than others.

On the subject of the big decision in the mines in Episode 2, our main change was in making Rosalina’s injury more graphically apparent — after the mine collapse, her legs were originally both buried under rock, which led to some players feeling like the consequences for saving Becka and Alexis were less severe than they actually were.

Before any changes

So we added some brutal clarity, that on top of making consequences more immediately clear, hopefully steered the scene away from feeling melodramatic (this teen asking you if she’s going to die with no visible injuries) to much more tense.

Brutal, but necessary clarity.

We don’t have new data on how that decision has changed for people, but there wasn’t too much else we could change (or wanted to change) about that decision — it will likely never be an even split, as balancing one kid losing a foot with two kids possibly losing their lives is an uneven bargain to begin with. But neither option is “good”, and the immediate consequences and regret are now there and won’t come as a surprise in later episodes.

Preliminary Results

Our preliminary results were mostly good, with most pick rates after the patch hovering in that 24–30% sweet spot. Keen Eye remained undisturbed, but Talk to Animals surprisingly lost a lot of ground. But then we saw Book Smart.

Oof.

Book Smart has been one of the harder traits for us to write, for a different reason than the others — it’s the closet approximation to actual player skill of our traits. Scarlet Hollow is a visual novel — a book, if you will — and drawing a distinction between the player’s ability to go through and comprehend text and the player character’s ability to do so proved difficult. This was particularly tough with the research segment in the library in Episode 2, where we felt like we shouldn’t limit players’ ability to engage with important books that held hints and details about the mystery based on their traits.

This led to Book Smart players feeling bad about their choice of trait, though, so we went back and added one more balance patch a week later, this time adding extra supplemental texts for Book Smart players to sink their teeth into, as well as gating some of the smarter reads on books behind the trait.

The result was a huge uptick of Book Smart respondents to our post balance patch survey, such that the results now look more like this:

Book smart jumping 5% in pick rate with only 56 more respondents puts its pick-rate post buff at 26.7%

On top of providing a better experience for our players, these patches mean that each bit of content we make is seen by more eyes, and they increase overall replayability.

If you’ve enjoyed this post, don’t forget to subscribe to this blog! You can also get early access to new posts (as well as to art from Scarlet Hollow, HD backgrounds and wallpaper and some of Abby’s comics) by subscribing to our Patreon. We’ve also got thriving communities on Discord and Twitter. And don’t forget to wishlist or pick-up Scarlet Hollow on Steam!