Designing for disaster in Finji's Overland

Overland, Finji's post-apocalyptic strategy game filled with many good doggos, landed last month on a whopping seven platforms: Windows, Linux, Mac, PS4, Xbox One, the Nintendo Switch, and the newly-founded Apple Arcade.

It's been in development for several years, while creator Adam Saltsman and his collaborators carefully worked out what would make for an effective and unique methodology for the game's apocalyptic design. 

After the game's launch, Saltsman dropped by the GDC Twitch channel for a brief chat about the design and development of Overland, sharing stories about the game's making and explanations for why it's turned out to be a somewhat grim vision. 

For your convenience, we've selected a few highlights from the discussion with Saltsman to aid you in your own game development journey, which you can now read below. 

Aiming for elegance in an apocalypse

Overland, like many games Saltsman dreams up, began in his mind as a series of screenshots. He first began to conceive of its design after playing XCOM: Enemy Unknown and Michael Brough's 868-Hack, an abstract turn-based dungeon crawler with a hacking aesthetic. 

"868-Hack is just elegant and compelling and beautiful and interesting, and I felt like it was getting at a lot of the kind of things that were exciting about XCOM," said Saltsman. 

"So the very first versions of [Overland] were kind of me trying to think about...if you mash these things up what do you end up with? and the original sort of doodle of like what a screenshot of this game might look like is like thinking of very old John Carpenter movies, and a handful of iconic characters out in the middle of nowhere at night trying to get some gas and get back in their car."

After dreaming up that Carpenter-esque vision, Saltsman said progress on Overland began to move forward after he brought on collaborators like art director Heather Penn, who began to take his turn-based, grid-based idea and boil it into a polygonal look.

But once the turn-based logic was in place, the question arose: what should players be actually trying to accomplish? "One thing that you run into if you’re trying to make a dense game, you know a game where you have a lot of interesting complicated choices but they all exist in a small space is you get kind of these information problems," said Saltsman. "Like how many hit points does that thing have, and what’s gonna happen if I do this? What are my options with this character now?"

"And a huge amount of time was spent trying to figure out what we wanted to do with that and what made sense for this game. There aren’t a lot of games that are squad based where each character has their own inventory for example, and in retrospect it seems like oh maybe there’s a reason that people were avoiding that. It’s really hard to do well."

Saltsman pointed out that in many strategy games like Overland (think Final Fantasy Tactics, the XCOM reboots, or other entries in the genre), it's difficult to look at a screenshot without having an intense knowledge of the game's systems and instantly have a read on what the situation is. Overland's design from there tried to evolve to make that possible. 

Saltsman used the board game Pandemic as an example. "Pandemic is this kind of like Risk except instead of armies its viruses you’re trying to control kind of these virus breakouts around the world," explained Saltsman. "These epidemics. And as soon as you know that the game is about epidemics, the basic game board is really communicative even if you don’t know a lot about board games."

"You can see that it’s a map of the earth. You can see that certain cities have a lot of cubes on them. You may not necessarily know right off the bat whether or not its good to have a lot of cubes, but you can certainly tell there are big differences between certain cities on this world map that you know, because you can recognize these places. They’re real actual places."

And so I think especially developing the visual look and feel of Overland was motivated a lot by those concerns, wanting a lot of the places in the game to be recognizable and for the stakes to be easy to understand and that all kind of is connected. I think to the way the gameplay has evolved, which was an ongoing, increasing interest in maybe initially making what we felt was a challenging strategy game but over time realizing that that was only kind of interesting because you were worried about these little characters."

Tuning for drama

For Adam and Rebekah Saltsman, a new phrase emerged for this design mindset: "tuning for drama." Saltsman describes it as a "ridiculous way of talking about this kind of game design," but it led to the Finji team moving away from strictly locked rulesets and into design logic that would encourage emergent storytelling. 

In one example, as Saltsman explained it, "in older versions of Overland the characters couldn’t move diagonally they could only kind of move orthogonally along the grid like most chess pieces or like all the monsters in the game. They can only move up down left right." 

"And that seemed like a reasonable thing to do. Everything follows the same rules. Cars go up down left right, monsters go up down left right, and characters go up down left right and that feels fine, cause you know you want consistency and you want fairness."

"But it turns out, it’s much more fun if the characters can move diagonally. You get like your scene here, like oh you get really cool close calls now."

Saltsman said this philosophy helped the team justify unusual and challenging design decisions such as the game's procedurally generated bios for its characters or a set of character eyebrows cooked up by Heather Penn. 

The goal was to keep enabling players to feel like they had agency and could orchestrate narrow escape after narrow escape. Even though that design meant tightening how much violence players could inflict on the game's monsters so they couldn't blaze through a level, the trend towards drama meant the game's difficulty was not so much about punishing poor decisions, but making the good ones feel all the more satisfying. 

A procedurally designed apocalypse

Overland's procedural spaces operate on two different levels. First, there's the world map, where players decide on which locations they wish to visit. then, there's the 9x9 grids, which are curated to create specific little spaces for players to build their dramatic escapes.

Saltsman dug into the methodology of the latter system, saying that he didn't know if it would work for other games but it seems to work for Overland. "Every object in the game, we give them—we call them placement masks--but they’re essentially gradients based on existing level features," Saltsman said. 

"The primary ones are where are the edges of the level and where are the roads. And we can sort of make a blurry gradient out of that information, and then start to overlap it with the placement masks that these different objects have."

"For example the placement mask for houses says you know what, these things don’t look good if they’re in the very bottom corner right there in the middle of the camera taking up the whole level. They also don’t look good when they’re in the road, and they don’t look great if they’re too close to the road. Something about it feels off."

"And so we can define a placement mask for the house that’s like you should be at least this far away from the road, and you should stay out of that southwest corner of the level. We can also keep track of how hilly the terrain is. And so we can have it check is this part to slippy or is this part kind of flat enough to believably place a human domestic structure." 

"So I guess the very first step is that the terrain that this level takes place on is actually quite large, this road goes on and on off screen both ways. We pick out the slice of it that we want to use first, and then we start populating it with objects. So once we’ve cut out that little square the little level generator will usually go through and roughly from largest to smallest will start to populate the level with key items."

"So there are many arrangements of roads and hills that will prohibit the appearance of a house entirely and so it simply won’t happen. And then it might go through— we usually prioritize your mission critical resources so this is I think a scavenging level probably? And so we’ll go in and place a few objects where you can go and open them up and get some items out of them."

"And each of those objects has its own placement mask. So smashed up cars usually show up near the road or on the road. Sheds usually show up kind of back towards the edge of the level with their door facing outward fences tend to be arranged along specific lines near the edge of levels that kind of thing."

Creating a very specific kind of difficulty

Saltsman admits Overland's tone and difficulty aren't for everyone. For every narrow escape, players are very likely to watch their survivors die out and leave behind poor, sad dogs that go out in a big explosion. (It's very sad).

Saltsman's dealt with feedback from some players that this situation is punishing because of the game's random number generation. He admitted those players have a point---the game's hard, no matter how you slice it, but when it comes to RNG, Saltsman has a very particular take on how games like XCOM treat their players. 

"There’s a lot of things in this game where you learn by doing. There’s not a huge formal tutorial process where you have to read lots of text before going out into the world. It’s a lot of sort of you know, thinking about what you’re doing, trying to make good guesses, and also paying a lot of attention to what happens," Saltsman said. 

This factor, and several others, are the primary reasons players might be frustrated or unsatisfied with the way their runs end. But Saltsman said it's more than the RNg that makes Overland challenging, RNG just is the asiest factor to pick out. "Let me put it this way--if your totally reasonable assumption on picking up this strategy game is that you’re going to kill all the monsters, get all the treasure and proceed to the next level, the RNG in this game probably feels very unfair," said Saltsman. "Because that’s not what you can or should do in this game."

Because Overland doesn't actually play out like XCOM or 868-Hack, Saltsman says it needs to take different methods to help players build a relationship with how difficult hte game is. "It’s funny because my relationship with those games is very much a relationship with trying to understand their approach to randomness. Like one thing we deliberately omitted from this game was percentage to hit. That was a thing I wanted to not be part of the equation for a wide variety of reasons."

"A big mechanic in Overland is that if you choose to fight off these creatures or you’re forced to fight off these creatures they usually make a ton of racket and it attracts kind of reinforcements. But reinforcements showing up is a big deal in this game, like that’s a really bad thing to happen to you while you play even if its frequently necessary and so there’s kind of this big stepped out delay there’s like two or three turns before the creatures actually appear in the level and you can actually see exactly where they’re going to appear ahead of time."

"We always thought of it internally as like proportional distance, so if something is going to happen that was gonna have a big outcome on the gameplay and it has a random element to it, that you get warning, that you get some kind of proportional warning in terms turns or steps or distance or something, right."

For more insights on the making of Overland, be sure to watch the full chat with Saltsman in the video below.