The Limits of Procedure

In a recent post, Tobold says:

After Everquest Next announced dynamic spawns with an ecosystem, destructible environment, and open world player housing, several people dismissed these claims as unrealistic with the argument that Ultima Online tried that and failed: Players kill mobs faster than they can reproduce, gather resources so fast that whole forests are turned into deserts, and build houses everywhere until no more flat surface is left. While that history is correct, the conclusion isn’t. As a counter-example look at Star Wars Galaxies, where open world housing clearly worked a lot better than in Ultima Online. So what is the difference, and how can a virtual world with a stable ecosystem be created without players destroying everything? The answer is simply in the size of the world, and the scale of the ecosystem.

Two paragraphs later though, he sort of undermines his own argument a bit (emphasis mine):

The Ultima Online ecosystem failed due to its small size and technical limitations of the time. But a procedurally generated world can be much bigger, even infinite. People don’t run out of world to modify in Minecraft. And if the world is big enough, you can put in enough mobs so that players can’t possibly make them extinct.

Er… if players cannot possibly make the mobs extinct, what’s the point in having an “ecosystem” in the first place?

But, never mind that. What I really wanted to talk about is the Minecraft bit. Because sometimes I get the feeling that “procedurally-generated” as a concept is being elevated as some paragon virtue of game design – a silver-bullet to boredom – when it is really nothing of the sort.

A procedurally-generated game solves exactly one problem: metagaming. That is not quite the correct word I want to use though, as you can still technically metagame a procedurally-generated world in several ways. Namely, by knowing its procedures. For example, no matter what Minecraft world is created (mods aside), diamond only spawns in layer 19 and below. Indeed, there is a whole bunch of Minecraft constants to consider (accurate circa 2010 anyway). All the Minecraft map being procedurally-generated does is prevent you from following a guide telling you to move North a hundred steps and then walk East to find a nice batch of Iron Ore near the surface. Avoiding those scenarios is certainly useful, and it’s always fun running into the truly bizarre and whimsical creations of the RNG gods.

The problem arises in treating the procedurally-generated whatever as “endless content,” when it is really not anything of the sort. As I talked about before, your knowledge of a game’s systems really only ever increases as you play; novelty has diminishing returns within the same game. Each Minecraft world generated may be completely different, but you are going to be doing the same sort of things each time: punch wood, make tools, make structure, dig for ore, etc. Different starting locations and details can spice up the gameplay for a time, but your “build order” is unlikely to change much no matter where you are.

Procedurally-generated content definitely has a place in gaming, usually in the context of roguelikes, but it will not solve any of EQN or any other MMO’s problems. Not that its application in MMOs made much sense in the first place – one of the defining characteristics of an MMO is a persistent world, which is incongruent with procedural generation. Even if it was just one world of infinite size (expanding at the edges), the amount of space surrounding objects of player interest would remain finite. And that’s assuming players would be willing to trudge out very far from their starting location in the first place.

No, just adding more space isn’t going to solve the issue. If anyone claims that an MMO’s world cannot be too big, simply point them at any given low-pop server of any game anywhere.


Posted on August 12, 2013, in Commentary and tagged , , , , , . Bookmark the permalink. 10 Comments.

  1. I think you’re selling procedural generation a little short. True using it for content generation can get boring quickly, but using it for map generation is totally viable.

    For instance, procedural generation is pretty much the sole reason any non-AAA game, like Minecraft, Don’t Starve, Dwarf Fortress, Elite, Rogue, etc, has large maps. Nobody would be able to develop such large worlds without equally large budgets were it not for procedural generation.

    Even on the AAA side it can help reduce the time and cost of rolling out new areas if used appropriately. Hopefully SOE won’t make a pig’s ear of the whole business in EQN.


    • Right, I’m not disputing the usefulness of procedurally-generated maps for regular games by any means. But in what context do they make sense in an MMO? Random dungeons, maybe?

      Then again, sometimes it just all seems superfluous. How many Minecraft or Don’t Starve maps are really necessary? How many maps does the average person go through in, say, the entire ~200 hours they might spend playing the game? You say these smaller games wouldn’t have large maps otherwise, but that’s only really true due to the metagaming concern. It’s a legitimate concern in the Wiki age, no doubt, but a randomized starting location on a much more finite number of large maps would serve a reasonable enough purpose for most players, IMO.


  2. I have to agree with Attic Lion.

    Procedural generation isn’t just to stave off metagaming (which it doesn’t actually do, by the way, provided you are using the same seed(s) as the person giving you information). It’s to introduce “emergent”/natural content without having to create everything yourself manually.

    While it is true that no procedural system can possibly “create” more content or information than is implicitly encoded into its own generation algorithms, those algorithms are becoming increasingly more sophisticated and developers are becoming better and better at layering procedural systems on top of each other to create better “hand-crafted” or “natural” content, as the case may be. Not only that, but it is entirely possible to write code which will vary your algorithms themselves procedurally, increasing the amount and type of variation you get in your output.

    You say that “build order” is unlikely to change but that is inaccurate. On a personal project I recently wrote code to procedurally generate a periodic table of elements before building a procedural “galaxy” built from the in-turn-procedural “chemical reactions” of those generated elements, ensuring that the very foundations of everything a user could experience in one particular galaxy would fundamentally differ from the experience in another.

    Procedural algorithms are much more capable than you give them credit for. It just depends on how good your algorithms are, and how far down the rabbit hole you want to go.


    • I have never seen the randomness or the “emergence” as a particular selling point in of itself though. It is just like all this virtual ecology discussion, going all the way back to Ultima Online: how or why exactly will it translate into fun? On an intellectual level, maybe, it’s interesting to think about. But you are only ever going to see or affect an incredibly small amount of these things. What’s the real difference between a procedurally-generated world and one that is crafted by an artist?

      If I play just the one map… nothing. Right? How many permutations would I have to experience before it matters that an algorithm did the heavy lifting?

      I’m not trying to denigrate the process here. I’m fascinated with the concept of emergent AI, which is essentially procedurally-generated behavior instead of maps. But sometimes a deeper complexity does not lead to greater nuance or even fun. That’s what we can’t lose sight of when we talk about these games. I’m sure there are technically hundreds of thousands of levels in A Valley Without Wind, but they really didn’t add anything to an otherwise poor experience.


      • The number of maps doesn’t matter. Nor does time played. I’m not talking about that.

        To continue using Minecraft as an example, if Notch and crew had opted to develop one really good map (and it would HAVE to be really good for the game to sell itself), say 25,000 sq ft in area they probably wouldn’t have been able to release an actual product for months, if not years later than they actually did and for greater cost. If they’d had enough cash to stay afloat that long at all.

        That is one of the great uses of procedural generation, to cut the costs of map development. It’s a great boon to games that are designed around using it. Or are on a tighter budget.

        Obviously in EQN they probably won’t do fully procedurally generated maps like in Minecraft, but cost cutting measures can still be implemented. Think of how many man hours can be saved by writing some rules to place art assets like trees, grasses, small villages, etc on a map instead of doing it all by hand. Even though they’d have to do a quality pass over the top of it it still saves resources once you have the system worked out.

        Is that an exciting use for it? No, not really. Hopefully it would at least lend itself to things like semi-frequent zone redesigns or swifter content releases. Plus the Landmark areas are almost certain to be procedurally generated as well just to produce the amount of space necessary for all those dick houses.

        Procedural generation really is more of a development tool than something for consumers to get excited about, but you know marketing tools and their love of buzzwords.


      • > It is just like all this virtual ecology discussion, going all the way back to Ultima Online: how or why exactly will it translate into fun?

        Hopefully in the form of emergent gameplay. You’re right in that if it isn’t engineered correctly it is largely a waste of effort. The difference might be compared to the difference between going to an animal park or a zoo and going on an unguided excursion in the wilderness of Africa. Sure, you might see the same animals and similar landscapes, but the potential for a raw encounter with the wildlife is only really present in the latter. It’s the UNPLANNED part of procedural content generation that lends it unpredictability, and it’s the unpredictability that lends it mystery, which can greatly raise player interest.

        When even you, as the system designer, can’t predict what will happen at the intersection of this system and that system under these certain conditions, that leaves the potential of a well-engineered system to create something new and exciting.

        > If I play just the one map… nothing. Right?

        Not necessarily, because you’re still getting a map that no or few other players would have had. That counts for something, right? Anyway, as you note, procedural systems go far beyond just map generation.

        > But sometimes a deeper complexity does not lead to greater nuance or even fun.

        I agree. Sometimes “procedural” is used as a buzzword and the systems aren’t well-made, but in such a case I would wager that same developer probably wouldn’t make a very compelling experience regardless. Procedure isn’t rocket science. If you’re gonna do it, do it right and make it worth your time and make sure your players can see the benefits.


      • That’s just the thing though: unplanned/unpredictable things occurring isn’t much of a selling point to me. “Emergent” is just another type of experience, not something greater or more compelling than if an artist/designer/writer created it by hand. You ask about the zoo vs a safari, and that’s a fair point. But I ask you about which would be more meaningful: a song written by an algorithm or a song written by a musician?

        Random maps have their place, and enable the exploration of game types that likely would not exist otherwise. But I am much more interested in having fun than the process of getting there, or what it took to create behind the scenes. And in the specific example of an MMO, I sort of mystified as to what applicability procedurally-generated content would even have.


  3. It will be very interesting to see what the final implementations are in game. I cannot tell if the MMO player base would even want the best portion of what procedurally generated content has to offer. From what I can tell most of the people I run into in MMOs want content that is focused around repetitive actions that offer a predictable path to properly itemized rewards. A piece of content could be the most exciting randomized thing ever created but if there is something next to it that offers equal loot in a shortened more linear sequence than you can be sure that is what groups will all gravitate towards.

    That said I am very glad EQ next is giving the procedural stuff a go. I suspect it will more help their dev team streamline creating new areas in game so that they can deliver content at a more rapid pace. It’s no silver bullet but cutting your design teams hours in half is a pretty big win for them.


  4. On of the most disappointing experiences for me in Skyrim was stumbling upon what I thought were more ‘free’ quests after finishing a particular faction’s questline and realizing that they were procedurally-generated and thus lacking reason, a story, to sustain them. Why should I care to clean up households of stray saber cats if there is nothing for me to piece together and become intrigued by, other than the gold I would be receiving?

    On the other hand, unassuming procedural generation can add a little bit of spice to the game, but ought to be used in the right proportions and areas. Dragon attacks, bandit appearances, etc. are randomly generated in this manner, and very entertaining. They could even be the cause of great anecdotes later, because randomness has a tendency to implode upon itself (a war between giants, dragons, and my horse).

    But when it comes to the terrain or the quests of my MMO, no procedural generation, please. I want to be able to recognize this area in whichever server I find myself in, and become attached to its crannies. I also want to observe the handcrafted detail that the artist put into a particular nook, because it is a significant location.

    Procedurally-generated signals to me that the particular area lacks a reason to be drawn the way it is. And I like to feel that the world has a reason to be how it is, much like I like killing old women in huts if the Night Mother commands it, and there is something to be learned about the contract. I don’t want to kill X first in Ashenvale, then in Barrens, then in Desolace, etc. for the same reason ([insert monster] are invading, Kill [number] of them!). I want to kill those X, gnolls, in Ashenvale, because they were planning to assault Astranaar, not because they were driven there by the previous horde that chased them out of Feralas.


    • Thank you! After a while, I start to wonder if I’m the only one being turned off by random quests, random mob placement, and so on and so forth. Those tools are fine in perhaps extending otherwise excellent gameplay, but I need more than simply pushing buttons; I need a coherent narrative, or purpose.


%d bloggers like this: