So you (along with several others, it seems) don’t see the absurdity of responding to “With work and innovation, we can vastly improve the implementation of X” with “But X has always sucked!”
Charles isn’t saying that procedurally generated levels are inherently superior and always have been, he’s laying out his ideas on how the process can be refined into a wholly new approach to the level design process.
While I WOULD say that procedurally generated levels have generally sucked I’m sure they can be made … okish, but doing so would require IMMENSE efforts that are vastly underestimated, imo.
What HumanTon referred to with his “General Theory of Crate Placement” is spot on. Coding a generation algorithm that can create consistingly convincing situations in randomly created environments is about as hard as creating an algorithm that can lead convincing conversations.
There’s SO much stuff to keep in mind, so many rules to formulate (and then obey), so many exceptions to rules that were exceptions to other rules in the first place … you get the point.
It can be done if you’re generating … say forests, or factory halls … or subway tunnels, but then the REAL problem only begins.
You need to code it so that stuff you want to happen occurs in a believable fashion, which I feel is almost impossible.
rezaf
And we’re saying that it would take a quantum leap forward in the technology to do it. A leap that would be more complex than just making the levels by hand instead.
Theres a reason noones done it and a reason it tends to completely fail so often. The Hellgate London mention is worth noting. The technology hasn’t really progressed since Diablo2 and not from a lack of trying.
Hmm. Maybe part of the issue here is that we’re just talking about different types of games. Maybe the word “levels” is misleading here because of course many types of games don’t really have “levels”. But here would be my list of some examples of games that are both good and have procedurally generated content:
Elite
NetHack
Stone Soup
Dwarf Fortress
Minecraft
Civilisation
Master of Orion
X-Com
Overgod
The Reliques of Tolti-Aph
Spelunky
…and that’s probably enough to make the point. Sure, the quality on that list is variable, but it covers a range of genres and eras of computer gaming, and it would, IMHO, be a pretty opinionated person who thought everything on it sucked.
Er… yeah lets see that work for the next Call of Duty. Or anything thats 3D really. Or set in a scenario remotely resembling the real world. It falls apart pretty quickly. Again, see Hellgate London.
I mean, in those situations it works fine, but those examples are pretty generic or abstract for the most part. It would never fly for a AAA title.
If, one day, somebody manages to attain this holy grail of lazy design methodology, I’d be curious to see how the same technology could be used to randomly generate good, compelling music; full of procedurally generated complex arrangements that denote not only a superior listening experience, but the artistic qualities of our finest composers and song writers – all while pushing the medium forward, rather than stagnating and serving up endless repeats of the same 4x4 beats.
Metallica would never have a bad album again.
Well, this is what I was talking about in post #40; FPSs would have to be the prime example of a mature genre. And indeed, “AAA” games these days generally means an FPS or otherwise another repetition of familiar gameplay mechanics given fresh appeal by the addition of new content. That’s fine for people who like that kind of thing, and I don’t mean at all to imply that enjoying such things is somehow bad. But I think it rather reinforces the point that there’s a correlation between mature gameplay mechanics and handcrafted content.
I’m not seeing how this changes anything then.
World-builder type games will be proceedurally generated, like they almost always are now. Everything else will be hand crafted because the mechanics of the game require it.
I mean as far as the list you gave almost all of those games just build a fairly generic world and then the players build on/change that world. Doing it otherwise is often very limiting or boring in those cases. But it wouldn’t work for most types of games, which is what the OP seemed to be implying.
This idea is all nice in theory, but in the reality of AAA game development hand crafted levels will always win out because there is no way to get procedurally generated content to look as good, and fantastic visuals will win any day over design. Not to mention it creates a lot of programming headaches.
As someone who has worked in both the RTS and open-world RPG genres, going to procedurally generated content is always the option of last resort, because it never is as satisfying from a player perspective, because as others have put it, it does lack a soul. It it an interesting way to discover your game though.
If you ask me the best way to do it for a AAA game is to make hand-crafted levels with procedural elements. That way you can have the best of both worlds.
And when I get down to those streets what will they look like?
What about when I go inside a building?
I can see your point of view based on the games you’ve worked on - AC is definitely more about the open environment with an over arching story that drives the player to specific locations. My experiences are more linear with games like Gears where the player is driven from A to B and has very polished experiences along the way.
Which, really, breaks down to the difference between open world and linear games.
I still hold that procedurally generated level shells won’t work for games like Gears. We put a LOT of effort into the shell of a level before any visual love or even lighting is applied to it. It’s like building a house - if the foundation is weak, the house won’t be stable and will have problems.
Now, talk to me about procedural systems within those level shells and I’m all over it. I would love procedural systems for things like adding decoration to an area … say, this area is in CityX so use this set of meshes and parameters and make this area look lived in for me, thanks! That would save a TON of time. :) Or buildings! Given a set of parameters and a volume to define the area, fill this space in with some sort of building that matches the theme of the current area and has some random elements of interactivity (you can go inside, maybe there’s a bedroom/maybe not, etc).
The concept of procedural content is intriguing but it doesn’t address things like hero pieces or specific spaces that the LD lays out to maximize gun fun. If you can combine the two into some sort of workable world building system, that would be something to see for sure. I saw a ton of places in Jamaica last week that would be amazing in a level but would also be a pain to build. If I could place down markers where I wanted a random shack generated that would suddenly become doable.
I can’t post links yet, but there’s a tech demo on youtube of a procedural city, complete with rooms and the ability to destroy things.
I agree that procedural wouldn’t really work for something like Call of Duty - it relies on being scripted to hell and I don’t think a computer could have the creativity needed for that sort of thing.
However, procedural generation should still have a place in first person shooters.
Multiplayer could get very interesting when none of the players “know the map” - you could discover new things every round. Procedural generation could provide a more varied setting, where map advantages are there for you to discover and use.
Admittedly, single player variations would be more of a niche thing. You could act as a soldier in simulated war between two AIs, fought over a procedurally generated world.
If a level is only going to be played once, procedural generation is not a good idea.
I’m not sure that’s anything but a completely trivial example. It doesn’t really matter if I get ambushed at step 3, or if I get funneled through passage X to get to step 3 and get ambushed there. Neither is dynamic, and even on a first play-through neither will seem particularly better. The real promise is instead being able to procedurally rearrange the story in reasonable, contiguous chunks: Say you break the story down into 5 sub-plots and they can happen in any order. Further, you can randomize something cosmetic about each (faction names, characters involved, etc…). Then you begin to cast the illusion of dynamic interactions when folks get together on a message board and talk about things.
Of course, this requires a lot more skill in narrative building because you have to design things so that these pieces interlock perfectly. The more you wish to present the illusion of dynamism, the harder this is.
Alright, thread listen the fuck up. I’m going to reiterate post 8 since apparently my english is too complex for people in this thread.
I am not saying you ship games that randomly generate their content. At all.
I am saying, you randomly generate the shell of the game, ONCE, and then the LDs populate it, paint it, add their unique flair, and set up the game within that world that now exists.
Aka, you randomly generate the first 80%, and then you can spend all your time on the last 20%. Then you ship what, as far as the gamer is concerned, is a static set of levels.
This has many, many benefits, not the least of which is that with a robust enough system for generating the world, you can very rapidly test out thousands of potential starting points, without ever bothering to do any actual work. You generate your shell, you give it a playtest – is there something you can use here? Yes? Great! Full steam ahead! No? Reload.
Compare this to a situation where you have a few weak level designers (because every team has them, I don’t care if you are the greatest game company to ever exist), and every time people aren’t happy with their levels, they have to go redo a non-zero amount of work.
It’s worth mentioning that this approach scales absurdly well. In fact, the larger your game, the more benefit of using a system like this. As I said, you can start with a game the size of Oblivion, rather than having to spend a year building Oblivion to start with.
In fact, in these situations, you can actually set up all your missions, plot, branching, storyline, and characters completely independent of the level creation process. With proper tools, the entire game can be built without even having a physical game world, and then you can generate the game world as soon as you like.
Consider the implications of this – you can ship a game with a much more polished and well built hand crafted system, but you can also ship a mode where it just auto-generates the entire game world. Replayability then becomes infinitely more interesting than just picking a different class, or a simple story branch. It would allow you to replay the game again for the first time.
Anyway, this is coming. AAA games are at record levels of unsustainability, and as more and more world interaction and world simulation become the expected part of games, it will have to be this way, or the entire industry will implode and we’ll be stuck playing XBLA games for a decade while it rebuilds itself.
The real divider is on whether or not you think playing XBLA games is a bad thing.
Also, aren’t they getting more expensive anyway?
It couldn’t possibly be that everyone understood you perfectly and simply happens to disagree.
Obviously you need to read post #8.
Daggerfall is what came to my mind as well when I saw this thread. I LOVED the random dungeons in that game.
I think there are folks talking about two different kinds of things here.
You’ve got a discussion about level design which tends to result in a static end product, the haggling is over the benefits of the handcrafted level vs. procedural design with later hand polishing. The other seems to be a discussion about the value of dynamic content and emergent gameplay. These are two different things.
Though I’ll add to lesslucid’s list of successful games with emergent gameplay The Sims and Civilization.