It would seem that procedural generation would be a great tool for open adventure games, since it would provide an almost limitless world. However, I notice most successful procedural generation games are of a totally different genre. The only game I can think of that uses procedural generation for exploration specifically is No Man's Sky, but it failed to meet expectations, probably in part because the procedural content, while vast, did not necessarily add as much value to the exploration game mechanics as people thought it would.

From attempts making procedural generation games and from what I see in other games, I see three problems hindering procedural generation in adventure games, and I cannot see any solutions:

  1. Procedural terrain may be straightforward, but procedural storytelling is not. I tried both. A few simple rules can make something that resembles a mountain or a city, but it is hard to make a game around it unless you have some sort of quest. It is hard to design a quest, though, unless you have a map to work with. How can you make a quest to find a place that might not exist?

  2. Procedural terrain provides scale without density. Sure, a procedural mountain might have all the textures and ridges to give it the complexity of a real mountain, but where's the fun it that unless there are lots of things to do on that mountain? Much of the strength of procedural generation is that it is easy to make something massive. If you make a world too massive without putting enough new things to experience in that world, it will just feel empty.

  3. The experience is not consistent. Much of the enjoyment from adventure games is reaching "that one part." Walkthroughs, tutorials, and wikis usually rely on the fact that players are all playing the same game, taking place in the same world. It is more difficult sharing your excitement for a game with others if the experience is so personalized. Imagine sharing a meme about the game if no other player in the world played a game with the same locations, characters, or creatures.

Keeping these problems, and perhaps others you can think of in mind, is there still a way to incorporate procedural generation into a game without ruining exploration? Maybe some certain limited procedural techniques that an adventure game could benefit from but not get in the way of a story?

  • 1
    \$\begingroup\$ "Bur it flopped" it sold roughly two million copies, on something like a 10-person team, which is pretty spectacular by most measures. It may have been poorly received in press & mainstream at release and lost additional sales to refunds, but it's maintained a core of extremely satisfied players to this day. It would be more apt to term it a cult hit. We should definitely take the lesson from it that it's easy to oversell "infinite" procedural content, and that communicating and managing expectations with players is critical, but to call it a flop is to ignore the very real success it achieved \$\endgroup\$
    – DMGregory
    Commented Aug 21, 2017 at 23:27
  • 1
    \$\begingroup\$ @DMGregory I would say it flopped towards the expectations everyone had before release. The game was still impressive for such a small team. But the overhyped release state and the false promises (Multi-player, content rich, greater complex story) broke it for the most people I would say. \$\endgroup\$
    – PSquall
    Commented Aug 22, 2017 at 0:21
  • \$\begingroup\$ Sorry if "flopped" was a bit harsh, I may go back and clarify. I was thinking more about expectations as @PSquall noted. It's an impressive game, but it seems like a good example of the difficulties of applying procedural generation to exploration mechanics since it was marketed as this game of endless discovery. \$\endgroup\$
    – tyjkenn
    Commented Aug 22, 2017 at 1:35
  • \$\begingroup\$ Rust has procedurally generated levels which you can explore about. \$\endgroup\$ Commented Aug 22, 2017 at 1:42
  • 1
    \$\begingroup\$ @Stormwind that sounds less like a comment than like the seed of an answer. ;) \$\endgroup\$
    – DMGregory
    Commented Aug 22, 2017 at 12:23

6 Answers 6


Procedural generation of environment is something which game designers experiment with for decades. There were some great steps forward in this area. Our toolbox of stock techniques grew a lot in the past.

But not so with procedural storytelling. This is an area which is still in a quite basic stage.

One approach of procedural storytelling is having a list of prescripted quest templates and spread them all over the procedurally generated world. Those templates usually have a number of variables which are filled in randomly from predefined lists or appropriate objects in the environment. Usually the player will soon notice these recurring patterns. At that moment the illusion of a living world breaks apart and all immersion is gone.

Another approach is to rely on emergent gameplay. Instead of pre-scripting stories, you design AI actors driven entirely by behavior rules. Those rules control all interactions with the environment, with other AI actors and with the player. They hope that this will lead to plausible social behavior which results in interesting stories emerging on their own. The main problem with this approach is that it is very hard to control and balance. The most interesting stories might take place in ways that the player does not even notice them. It can be very hard to get these AI actors to perform these interactions in the way that it makes for the most interesting gameplay experience for the player.

One game which does this a lot is Dwarf Fortress. Does it have interesting memes for the community to talk about? You bet it has! Interestingly, most of these were born from glitches and quirks of the emergent gameplay systems. Among these are the horror of killer-carps (AI-controlled fish which were far too aggressive and overpowered in early versions), Catplosions (population explosion of cats which was hard to prevent due to dwarfs being very attached to them) or the mantra "Losing is Fun!" (a way to cope with the frustration which can result from the emergent gameplay generating unwinnable situations). So your fear that players won't have anything to talk about when everyone's experience is different is quite unjustified. To the contrary: Emergent gameplay gives the players even more to talk about, because they want to share the unique and interesting stories they experienced.

  • \$\begingroup\$ Good point about emergent gameplay. That may be something I need to research more. I've tried the template approach and had the same problems you described. My hesitance in using emergent gameplay was the difficulty in implementing and the inevitability of glitches and quirks, but you bring up a good point that the latter may be a non-problem. \$\endgroup\$
    – tyjkenn
    Commented Aug 22, 2017 at 19:43

I think a good example of a procedural "adventure" game (by some definition), one where the procedural content promotes exploration, is Terraria. I think it handles many of the pain points that you outline quite effectively:

Procedural terrain may be straightforward, but procedural storytelling is not.

"Storytelling"? Terraria doesn't really do that. But it does have progression and de-facto quests (I'm not talking about fishing quests). It accomplishes this through a combination of factors.

Progression is not usually based on getting to a single, specific location. Terraria is a game about stuff, so progression is based on getting stuff: harvesting resources, opening chests, killing monsters for loot. Of course, resources, chests and monsters are all based on locations, but this is done in a very broad way. There are large biomes, and throughout these biomes, you'll find whatever resources are specific to them, randomly placed throughout.

There are elements of progression that are based on specific locations. But those are always generated in "specific" locations. Floating islands are always floating in the sky. The Dungeon is accessible from the surface on one side of the world, and the Jungle is on the other. Corruption/Crimson areas are many and it's pretty much impossible to miss them (unless the RNG was especially bad to you).

The least consistent place in Terraria that is important for progression is probably the Jungle Temple. And even that is not something you're likely to miss.

Procedural terrain provides scale without density.

That all depends on what you generate. Earlier versions of Terraria had problems with this. The underground areas could become very "samey".

Terraria, especially 1.2 and 1.3, solved this problem, making them incredibly dense with stuff. There's always something new to stumble across: ruined houses with chests, icy caverns, abandoned railroad tracks, underground lakes, underground desert, marble caverns, etc.

Density is created in Terraria through diversity. Instead of repeating the same content over and over, you repeat new contents. You have enough variations of content that the player doesn't get too bored.

What Terraria in particular does is that it builds the basic terrain of the world. Then the world engine goes in and randomly puts things in various places. A marble cavern here, a ruined house there, a shrine over there. The icy caverns go over there, a big desert sits here. The system is designed to be able to slot these features in anywhere in the world. That ensures that there is enough density of them that you won't get bored.

Now, the world engine isn't stupid. It biases certain features based on depth or distance from the center of the world (where you first spawn). But the main goal is to make sure that the game is filled with lots and lots of new stuff.

It may not make much sense why there's a ruined house perched above a lake of lava, but the player will appreciate that it's not just another cave.

However, diversity can only be achieved if the experience of play allows it. Terraria is an open-world game in which you can build lots of stuff. But it also has an action-oriented combat system that offers many dimensions of possible play. There are four "classes" (defined by what bonuses the armor you're currently wearing gives you): melee, ranged, magic, and summon. And within most of these classes, there is a large diversity of play options.

Ranged characters have a number of weapon types, but they also need to deal with the fact that they can't take a lot of hits, so they need mobility items. Melee characters need items that improve their ability to take hits. Magic characters need items that deal with mana regeneration. And so forth.

There are two important elements here: the diversity of play experience, and the lack of an enforced class system. The latter is really important, since it means that if you find a good melee weapon or a good ranged weapon, you can still use it even if you're not in ranged or melee armor. You may not be as effective as a dedicated build, but it's not useless to you. You can always try it out and see if you like it.

Plus, this means a "class" change is just an armor-switch away. If you get a good melee weapon, you can probably craft some melee armor and try out a melee build. If it doesn't work out, switch back to your ranger.

In this way, Terraria creates a game system where the diversity of exploration encourages a diversity of play, since there are a lot of options in the play experience. And vice-versa: diversity of play is what makes the diversity of exploration possible, since it allows the creation of a multitude of items.

The experience is not consistent.

This is where diversity and scale meet. Terraria achieves a surprising level of consistency in its overall play experience. While each individual world is different, the scale and scope of play means that you're unlikely to come up short on something important.

You're going to find enough Gold to make Gold Armor, if that's what you're interested in. You're going to find enough cobwebs to make spellcaster robes, if that's what you're interested in. Etc. All of the essential stuff is available in sufficient quantity. And thanks to the diversity of stuff, if you don't find one particular kind of thing, you probably ran across something similar to it.

Furthermore, there are a number of checkpoints you have to cross in order to move to another sequence of progressions. Surviving the Underworld is extremely difficult to do unless you have access to the Jungle/Dungeon/EoW/BoC-teir of equipment. You cannot shift your world to Hardmode unless you fight a boss in the Underworld. You cannot harvest Chlorophyte from the Jungle until you've killed all three of the Mechanical Bosses. And so forth.

Because of these checkpoints, you're encouraged to gather more resources until you can pass them. Which ensures that you have the chance to run across lots of various loot, thus ensuring a relatively consistent power level.

Now, that doesn't mean that inconsistency doesn't happen. Bad world creation can put you in a tight spot sometimes, or maybe you find yourself in a boring patch of world. But this is quite rare, thanks to Terraria's massive diversity.

The worst the world creator will do to you is feed you lots of the same resource.


As you said, procedual generation is a good tool to create open world's with a lot of possibilities. The obvious downside is, that it is hard to balance in many regards: either it feels to empty or to cramped, to "over the top" or dull...

Good adventure games live by the atmosphere, especially those with a big open world. My favorites in this regard are Gothic 1 and 2. You could never recreate this atmosphere with just procedual generated environment, I guess.

The only way to work on your points is somehow circumvent them.

Procedural terrain may be straightforward, but procedural storytelling is not.

You could go the Minecraft way of creating biome. Maybe a city biome, a hill biome, a desert biome. Then fill the city biome with a "pre-designed" city structure, add important places, then add quests and bind them to a hill biome nearby, connect those via streets and add the needed dungeon structures to the hill biome. This way you can add other random structures around the biome, maybe with possible loot, to encourage more exploration. Technically it's still procedually generated, but you need a lot of work for the structures, quests and the assertion that all needed structures in every needed biome is generated. Starbound does something similar, where you have to gather resources from other planets, or it creates an instance of a dungeon to explore.

Procedural terrain provides scale without density. 

Imo Skyrim got a lot of "empty" places. Just walking around from town to town took a long time, but you always had something to do or look at. Beautiful flowers, mountains or epic buildings, then gathering stuff along the road, kill the wildlife or discover another empty hut. Yes, Skyrims world was designed, but with enough content you can encourage players to walk of the tracks, just with randomly placed structures, enemies and just stuff. Again, I think it's hard to balance, but I would prefer cramped world's over empty world's, at least partially.

This is actually what I think No Man's Sky did wrong. I have not played it, but watching it was boring. Walking around on the planet, everything seems lifeless. You had goofy animals walking around, maybe 4 of the same kind, some resources, rarely some technology or alien words. That was it. No really big structures, big groups of animals and no interactions between them.

The experience is not consistent.

Yes, and that is most likely unfixable. As I said, atmosphere is what carries an adventure game. On the other hand, players love to enjoy things that are unique to them. Maybe one player has to fight a boss in a room with a complicated layout or other enemies from a nearby dungeon can rush in, while another player was able to cheese the shit out of this exact same boss. These might be things players want to share or brag about. It might also encourage creativity.

  • \$\begingroup\$ I thought of using something similar to Minecraft biomes. The problem I saw was that people mostly play Minecraft to build, not explore, and the procedural terrain acts more as a foundation and as resources for building. However, you do bring up a good point with the idea of using pre-designed cities. A procedural patchwork of pre-designed structures could provide quests that still make sense. \$\endgroup\$
    – tyjkenn
    Commented Aug 22, 2017 at 20:09

You misunderstood procedural generation. It's not about slapping together a noise and a mesh generator. Most successful procedurally generated games used some sort of pre-designed architecture.

Let's look at some examples. First of all, Spelunky. It's praised by almost everyone and it used procedural generation. It worked with a 4*4 map, it chose a path first from top to bottom, then filled them with premade rooms and made small changes to them. This made the game feel fresh and fair every time, but it was always completeable.

Another example is Cube World. It mainly used a (low frequency) noise generator, but it also added pre-defined features to the world, such as huge monolith like cliffs and variois buildings.

If you want a game with a story that used procedural generation, then just look at the first four Elder Scrolls games, especially the first two, Arena and Daggerfall. The designers created every important dungeon by hand, but the other dungeons were randomly generated.

Yes, the experience won't be consistent, but it isn't consistent in other games either. Most games will have better and worse moments.


The complexity of managing procedural world generation and play-balancing can outweigh the benefits. The major early benefit to proc gen was to save developer time over the long haul; only in recent years has it come to be lauded for attributes outside of that (though Elite and Captive were great early examples of achievable playability, even they were not purely procedural).

The best way to manage complexity is to build tools to do that for you. We currently spend a lot of time manually fine-tuning parameters to allow us to build the procedural worlds we want. New biome, new parameters. New quest/story type, new parameters. What becomes clear only after a much longer time of building procedural worlds, are the patterns behind these processes which then allow us to build tools to amplify our work rate. (This is true of any area of development.)

Since we all develop differently... I'd encourage you to try it the manual way enough times that those patterns start emerging for you, for the problems you often have to solve. Once they do, you'll be on the road to building tools to support that work. Really, it's fairly bleeding edge knowledge generation.

Those parameters you mention, scale without density, inconsistent experience, can all be addressed by proper (runtime or compile time) automated tuners. The complexities of story management can be handled by a combination of a sound knowledge of graph theory and narrative theory.


Procedural terrain may be straightforward, but procedural storytelling is not.

Procedural terrain obviously gives you terrain to work with. This is not bad and it's not the only thing you can generate. You can generate dungeons ala Roguelikes and you can generate Cities if you use rules similar to City Building games like Anno or Impressions Games.

But terrain is not gameplay, so you need to add the proper games.

It's actually not that hard to generate a dynamic living world.

X3 Albion/Terran is the best example of this. It has trading gameplay with even building stations to get resources and create products, it has a space combat game where the player pew pews with pirates and it has a bit of a Strategy 4X flair with faction conflicts and bigger fleet battles that works pretty well with its economy.

How can you make a quest to find a place that might not exist?

Procedural generation can give categorized data that you can use. When it creates a cave or a mountain it can tag that area as such. This could be used in a quest to populate that area with a monster. This is how Dwarf Fortress Adventure mode generates its quests I believe. In a Minecraft like jungle biome you can generate a special herb used in a quest. In an ancient ruin dungeon you can have a special artifact for a quest.

Procedural terrain provides scale without density.

If you want to go beyond procedural generation then you are looking at user generation.

Imagine a world where players can create their own personal empire with Factorio style in depth development and logistics as well as 4X style research.

Now imagine all players uploading their empires onto a shared world that they can explore.

People can explore the ruins of player empires and their technology.

People can even fight with their empire against a AI controlled uploaded player empire.

A True Exploration Game is User Generated.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .