Tides of Chaos: How Baldur's Gate 3 tackles the RNG of D&D-inspired dice rolls

This 2020 interview covers the Early Access version of Baldur's Gate 3, and some aspects of the game may have changed since its publication. For a closer look at the full release, check out our dive into what Baldur's Gate 3 does (and doesn't) get right about D&D-driven systems.

Larian Studios landed the monumental task of picking up the long-dormant Baldur's Gate series, a project that tasks the team known for its own recent fantasy RPG series success Divinity Original Sin with both reviving a decade-dormant franchise and working the mechanics of the TTRPG Dungeons & Dragons into a game for modern audiences.

With the game now a week into its Early Access cycle, Larian Studios lead systems designer Nick Pechenin sat down with Gamasutra to discuss how the early days of that open development process has gone and offer a D&D-driven dive into how something as core to the tabletop experience as rolling a twenty-sided die becomes ever more complicated in the context of a video game's narrative, combat, and more.

The following Q&A has been edited for length and clarity.

So I guess the first question would be congrats on Early Access, and how is that whole thing going now? Three days in, right?

Pechenin: Yes. Three days almost exactly. Actually, it's been way bigger than we were expecting. You might have seen tweets by our founder Swen Vincke. That it's, it's been phenomenal. We're seeing so many people playing, we're seeing a lot of data come in. And it's very interesting to see what players get up to. It's very cool to see them spend so much time in character creation, which we tried to polish up as much as possible. So it's been a blast. And our kind of worst fears have not shown themselves while our biggest hope, that we will pull off something that feels like D&D, has come together.

Is there One Secret Trick to pulling off Early Access? What's your one big piece of advice on something all studios attempting an Early Access campaign should make sure they're doing when they take on that process?

My advice would be to be prepared to throw away your design docs on at least some of the features and to respect the process. I think if the studio is doing Early Access, they need to realize that this really means that the players can give you feedback that does not align with your design vision at all. And you will have to reconcile that. But doing that will actually make the game better for a lot more players as it turns out.

So my advice would be to be flexible, design wise, and to plan your design ahead in a way that is kind of modular. This is something we got a quite a bit better for this project I think. We have plans for systems that we foresee that we'll have to add due to new classes, new features, things that we haven't talked yet about. But we try to make sure that all of them can be swapped for something else that the players will say is really a priority for them, or can be changed in a way that will not make us remake everything from from the get go. So we try to keep systems kind of modular. Flexibility is really the key here.

The scale of Baldur's Gate 3 and, I suppose the size of its playerbase, is already much different than Early Access with Divinity Original Sin 2. With that in mind, how has the Early Access process been different between the two games? What lessons have applied to both?

Something that we took from the previous Early Access was that being upfront and as transparent as possible really pays off with the players. It's very important to set up expectations early. So I hope we're trying to be very, very clear that this is going to be Early Access, there will be issues, we will try to fix them as fast as possible. But indeed, this is not the final release yet.

Another thing is when you're designing a game, you have a lot of ideas of what you want to put in. But only after listening to a lot of people you will realize that some of those ideas are your darlings, and that maybe they should be put aside. Then there are issues that you never thought would be a big deal but you really should pay more attention to. So what we're seeing right now, for example, is that we're starting to discuss more of the nature of randomness and how we approach that.

And because the audience is so much bigger and they're coming from such different games, their expectations for how narrative is handled is very different. So for a D&D player who is completely cool with getting three ones on the D20 in a row, they go like "yeah, this happened yesterday to me, this is completely normal." And then there are people who are coming from titles like XCOM or something more strategic where they would expect some dampers or stabilization on RNG so that you never see a really bad streak or something like that. So we're now discussing how we're going to tackle that.

There's this occasional dishonesty to randomness in games where, behind the scenes, some games will lightly lie about percentages and treat numbers over 90 percent as a guaranteed hit, or below 20 percent as a guaranteed miss. But for your team, are your dice rolls actually just a random one-to-twenty number generation or is it weighted to give more balance and fairness?

So what you're seeing right now is as random as computers can possibly output. It's just true randomness, and we're seeing that it works for some players and some players would like more options and more control over what's happening.

What we're seeing also is that this pure randomness works for some types of rolls: if it's a smaller situation, or it's a very long shot for you to do this, it's okay for you that it's a really random roll. But then if it's a huge choice, or a huge chance for you to change what's happening with your party or with the characters you're trying to help, you really want to have like a little bit more control than just, "here's the dice, that's how it's gonna go."

From our own experience at the tabletop, we know that even though dice are supposed to be pure randomness and very honest, the DM has has the screen for a reason where they're rolling their dice in secret. That is already in D&D this built-in mechanism for stabilizing randomness and an understanding that creating a compelling narrative takes a bit more than just completely [rolling] in a random motion.

It has to be handled very carefully, because players are very good at spotting the game putting its thumb on the scale and the cheating the randomness. So right now we're discussing where exactly we're gonna start stabilizing RNG, most likely in combat scenarios. This is something that people have very specific set of expectations for. It's where they want a lot of control, have a lot of plans, and come up with very interesting tactics and strategies. If you have too much RNG it just messes it up. It devolves tactics to something less interesting. But when it's about narrative, we're going to be looking at what types of rolls are okay to keep random and what types would need a bit more control.

What's Larian's official stance on save scumming for better rolls?

Some players want to save scum! They feel like this is the degree of control they have over the experience and they want to exercise that control. My fiance, she loves to save scum and if the game will not let her save scum she will make sure that she will hack it to be able to. She will crash it through ALT+F4 if the game is trying to save on quit or anything like. So if players want to save scum they're gonna save scum. And then if a significant portion of your audience wants to save scum, why make it less comfortable for them?

I can't remember the game name but I read an interview recently where they talked about how their randomness was pre-determined, so like your next three random number rolls would be set in stone. So even if you reloaded a save it would be the exact same results because they wanted people to be locked into the chaos of the universe.

I think actually we had a glitch in Divinity Original Sin 2 that accidentally introduced what's called random seed fixing, where when you reload, it's the same result. We saw a huge backlash, even though it wasn't even intended. It was a bug, but immediate, huge backlash because in a way, when you do that as a developer, you're sending out the message that there is fate and everything is predetermined in a way. Players really negatively react to that, or at least enough of them for us to not really consider that.

But on the kind of opposite side of the spectrum we know that there are players who want to have a true D&D experience where you roll with the loss. When you fail at something, you just go over it. They want the game to help them to avoid save scumming. So if save scumming is too easy, they might want a mode where save scumming is not available through some mechanism. They agree to this rule with the game and they try their best to go for the game without save scumming and see where that brings them. So this is something we're also discussing how we can introduce that.

How do you balance the ability to fail checks in dialogue or exploration with still wanting players to progress in the story and get meaningful information as they do? Can you talk a little bit about how the team went about making that system in a way that gives players flavorful options despite opportunities to fail rolls?

Whatever flowchart map Larian has must be massive from just like the sheer amount of skill checks and class/race/background driven dialog options.

The dialog trees are absolutely gigantic when you start playing them out. I hope maybe we'll share some screenshots later in like a postmortem or something. But they're gigantic. Like, if you look at one, you can't possibly read what is written there without zooming in.

What we tried to do with the challenge and the randomness there is that we try to keep the difficulty check numbers in certain buckets so that the player can get used to a certain amount of randomness in some situations. (Ed. note: DC is Difficulty Class or Difficulty Check, tabletop slang for the number a dice roll has to beat to be successful.) So if like an easy check is 5 DC and a normal check is like 10-ish DC, the players who are not even familiar with dice rolling will start to kind of get an emotional response to what the certain DC difficulty check means for them. For them, gaining bonuses for that or trying to game these checks through drinking potions or finding items that give you bonuses, they will see very very clearly that they're getting those bonuses.

But what we also want to do, and that became very clear when we saw people actually interact with these rolls, is that we want to change how they're displayed. So right now we try to have a minimalist UI for the rolling. You see a number that is directly comparable to the dice. So if it says 10 you really need to see 10 or more on the dice to pass a check, right? But this 10 is actually not the DC; it's actually the sum of all the bonuses or penalties you might have.

So there's a lot going on there, and we even called it the "target" instead of "difficulty check". But we realized that we actually have to show you a bit more about what's going on there's so that you realize how this number is made up. Then you start having ideas for how you can master this number, how you can manipulate the chances in your favor.

We keep asking playtesters and we keep looking for players trying to manipulate the chances. There are a lot of ways in the game to do this, but they're not tutorialized currently. Savvy D&D players know about them and utilize them. But we, of course, want to teach everyone to do this. And I think clearer UI feedback for the kind of bonuses you can have is going to help players start getting ideas of how to manipulate.