Tales of the Rampant Coyote

Adventures in Indie Gaming!

Procedural Content: When it sucks, when it doesn’t

Posted by Rampant Coyote on August 27, 2012

I’ve been a fan of procedural content since I first played Telengard back in 1982 or so. The Telengard dungeon boasted a size that was pretty much unattainable by conventional methods. There were a total of two million rooms in the game. Even if you managed to compress the entire room information down into a single byte (I don’t think that would have been possible with the room data variations, but two bytes would probably have been attainable), there’s no way you could have stored all the rooms into 64,000 bytes of memory on the C-64… or a 170k floppy disk.

But while two million rooms sounds cool on paper, the reality was kinda boring. Worse, it was frustrating, as a teleport trap required you to pretty much scrap your old map and start over again, because there was simply no way you were going to find your way back “home” again short of a recall spell (which would require a sacrifice of your gold, which represented a pretty good chunk of your experience points).  Nobody on earth was going to explore all two million rooms, and with only a handful of features to differentiate them, one collection of rooms didn’t seem that different from any other. Encounters, treasures, and responses to actions (what happens when you sit on the throne?) were all random, so very little of it mattered. All that did matter was learning where the stairs were.

Other games had far more interesting procedural content – like Frontier: Elite II, where a neighbor of mine and I would keep and share records of the best trade circuits and areas in the galaxy with the most interesting or profitable gameplay. This felt far more like “exploring” in the classic sense. It was the same idea as Telengard – with a similarly impossibly huge, fixed galaxy – but it was much more meaningful. These weren’t just places, these were things I could use.

Back in the 1980s, I was introduced to the game “Warhammer 40,000.” My friends spent horrendous amounts of money on miniatures for that game – a temptation I narrowly avoided after a couple of play sessions that exposed me to just how fun this dark science fiction wargame could be. The first time I played, a friend acted as a referee between my brother and me – neither of us with any experience in the game. My friend pulled out some plywood boards that had some blocks glued onto it in some resemblance of a city-scape. Then he pulled out another bag of oddly-sized wooden blocks, and dumped them onto the boards. “This is your battlefield,” he explained, “the ruins of a city. The blocks are buildings, some of them collapsed. You start on opposite ends of the map. Last one standing on the map wins.”

For movement, distance, and line-of-site calculations we used a piece of yarn. If you could stretch the yarn between two miniatures on the map without touching or bending around any of the blocks, you had line-of-site. To determine distance, you’d measure the length of yarn between them. It was very organic and fluid, and my squad of space-orcs enjoyed a tremendous victory, so it was awesome. The random terrain generation was a hit with me.

Ditto with X-Com, which I played around the same time as Frontier. While there weren’t many variations in battlefield layout – it all looked the same very quickly – the trick was that it was all useful. The positioning of the farmhouse, those awful grain fields where aliens always seemed to be hiding (one incendiary rocket to perform “reconnaissance by fire” was always useful… if you hear an alien death-scream in the middle of the flames, then you know there had been an alien hiding there).

Diablo-style and roguelike games, however, never really thrilled me with random dungeon generation. Like Telengard, the only really interesting part is finding the exits. Otherwise, it’s just background scenery. It doesn’t feel like exploration because in spite of the most interesting dungeon generation systems, it’s just not interesting because it doesn’t make much of a difference. It’s just a place where occasionally interesting things happen. Now, the generation of treasure is a whole ‘nother matter. Again, it’s a random thing… but makes for extremely compelling gameplay as every item represents a potential upgrade to the player’s abilities. While the vast majority of items are simply tossed into the inventory to be converted into gold at the first opportunity, the promise of meaningful upgrades – which may occasionally add such new abilities to the player character that they change the feel or pacing of the game – give this form of procedural content generation an almost addictive flavor.

Minecraft is fascinating with its random, procedurally-generated worlds. Again, it’s not just the cool vistas and bizarre landscapes that it presents – but the importance of that exact configuration to the player. If anything, Minecraft is perhaps a textbook example of procedurally generated worlds done right. When you are dropped into a new world for the first time, you are immediately confronted with the need to survive and very little time to make preparations against the monsters that come in the night. Every tile in the world can be important in this respect, and you must immediately assess the value of the nearby resources and locations to build your shelter. Later in the game, the search for rare materials and the even rarer dungeons turns the entire world into the equivalent of Diablo‘s random item generator,  as every dug block or uncovered cavern yields the promise of finding a vein of useful materials.  And of course, there’s the pleasure you get by playing god with the terrain and shaping it to meet your own needs, desires, and whims.

There are countless other examples, but for me, what’s more important than the cleverness and variation of the procedural content generator is how much it influences my gameplay and decision-making. An interesting battlefield with terrain that can be used to your advantage is always more interesting than one that is simply scenery.

This principle isn’t entirely restricted to procedural content.  While hand-generated content may have some cool signatures of human creativity and a stronger sense of purpose that makes the world feel more alive, it’s still vastly more interesting to explore when it has a purpose beyond mere scenery as the player progresses from point A to point B. Exploration – to me – is about trying to discover cool and interesting things. That usually means something that influences story or game mechanics, or gives the player a chance to learn something new about the game world, or provides interesting choices or problems to solve. Otherwise, it’s just filler. And I really don’t want to play a game that’s mostly filler.

Filed Under: Design - Comments: 5 Comments to Read

  • Adamantyr said,

    Wow, strange coincidence, I just started working on a Roguelike for my vintage TI-99/4a… which oddly enough no one really took the time to create.

    I agree on the “sameiness” of procedural content; I looked at the algorithm used to generate Rogue maps and it’s way too predictable… 9 rooms with connecting corridors. BORING.

    My plan is to investigate three different systems:
    – A grid-based room plot system which plans out connections
    – A geomorph design with room/chamber designs stored on disk and plopped in and rotated as needed
    – A “drunkard’s walk” system which basically starts on the map and just moves about plotting things, much like a human DM with a piece of graph paper would

  • Craig Stern said,

    I agree 100% with this. I did a short write-up about an early version of a tactical roguelike called Dungeon Bash Tactics, and one of my critiques was basically what you write here: there were no terrain effects to give consequence to the variation.

  • Maklak said,

    Anarchy Online has random missions. They are instanced dungeons in pre-defined locations (hundreds of old buildings, caves and so on). I don’t think, I’ve ever seen two identical missions, but they do get repetitive after a while. The missions were made from pre-defined rooms with doors. The rooms were divided into themes, such as cave and lab and a mission always has a theme. The entrance to a mission always has a small safe room (well, unless mobs in the next room wander close to the door, open them, see you and attack) connected to a corridor. The doors in the corridor lead to other rooms and corridors. Each room has 0-2 enemies and 0-2 chests (some of those being junk piles, old skeletons, but they work like chests.) Of course this led to room layouts that made no sense, but even manually designed dungeons seldom make sense.
    Some people complained about missions being boring and repetitive, but I like that part of AO. There are always some variations. In any case it is a much better solution than having the same dungeon for every mission, like in Mass Effect.
    I believe Diablo uses a similar system of random dungeons and Din’s Curse does too.

    I play Dwarf Fortress from time to time. It has a lot of randomly generated content and some people enjoy it a lot, but I don’t care about most of that stuff. There are always some civilisations, cities, randomly generated beasts, such as giant birds made of vomit, geology, but only some of these things matter and some kinda even-out with the law of large numbers.

  • randomelginguy said,

    You should check out Spelunky for XBLA or the free PC version, it’s a platformer that has roguelike elements, it’s the level layout that’s random, I thought it was quite good.

  • John Evans said,

    It might be illustrative to look at How to Host a Dungeon – http://planet-thirteen.com/Dungeon.aspx . It’s a set of rules for procedurally generating your own dungeons “by hand”, with paper and colored pencils.