The Uniqueness of Worlds is About to Get Nuts
One of the biggest draws to the initial design pitch of Crowfall was that its many worlds would be unique - no two sharing the exact same layout and peaks and valleys. Well, that idea is about to come to fruition in the alpha, and we spoke to Todd Coleman about what it all means for Crowfall moving forward.
MMORPG: What's taken so long, gawd... no seriously, what was the missing piece?
Todd Coleman: Wait -- are you asking "Why did it take you guys almost two years to build this", or "Why have I been playing MMOs for 30 years, and I've never seen this before"?
In both cases, I think the answer is the same. This is one of those incredibly challenging technical hurdles that can swallow up years of R&D effort.
Procedural generation is powerful, but it's also incredibly hard to control.... it's more like a nuke than a sniper rifle. Massively effective, yes, but you don't have a lot of fine control over the result.
We tried to do unique worlds on Shadowbane back in 2001, and after 3+ years of engineering, we had to fall back to a single map for all servers because while we could make these enormous worlds almost instantaneously, they weren't particularly playable. The goal is not to "make a world that looks cool", it's "make a world that looks cool and results in playable spaces that feel engaging and fun."
Our solution, which seems to be working really well so far, is to build "parcels" of land that are hand-crafted (think of them as terrain segments shaped like giant Tetris pieces), and then stitch those pieces together seamlessly to create an endless number of unique worlds. All of the content within each parcel can also be defined, so that every "L-shaped, 4 square swamp" can be different.
MMORPG: How does the procedural generation of each map work? What makes it so you have control over how things are laid out?
Todd Coleman: Well, hold on -- the system isn't done yet! We hit a major milestone, but there is still a lot to do.
Technically, there are two parts to this: the underlying parcel system, which allows us to stitch together land masses into a seamless whole AND a procedural system that automated the process of laying out those parcels and all of the content that fills them.
The first part, the underlying tech, is now up and running. We can build any number of unique worlds and boot them and players can log in and harvest resources and fight monsters -- it's a fully-functional MMO.
The second part, though, is still heavily dependent on designer input right now. We have tools to configure everything, but someone has to walk it through the process, mixing and matching the pieces and reviewing the map and changing the default content for each parcel.
The next step will be automating that. Our plan is to lay out the major terrain features (mountain ranges, valleys, hills, forests) and then have a "history simulator" for the world that fills in all the details. The simulation will start by populating the world with various races, and then use AI to let them explore and conquer the world around them, sort of like a game of Civ, only completely automated. As the races explore and capture resources and bump into each other, they will go to war, declare peace, build cities, etc. This process will populate the world with ruins (and history, and artifacts) for the players to discover once that world comes online.
MMORPG: What about clipping bugs, things that shouldn't be on a map, etc. Do you guys "spot check" them before people play on them?
Todd Coleman: So far, the system has been pretty resilient because it's not purely procedural. The parcels are created by hand, so we can test each parcel to make sure there aren't clipping issue or bad slopes. We've had a few bugs pop up around the seams (where the parcels are stitched together), but most players can't tell where one parcel ends and the next one begins.
MMORPG: When worlds "die", eventually, does that mean we'll never see a world again? Or will you possibly bring it back for a special campaign later?
Todd Coleman: The world goes offline, which means that specific campaign is effectively over. We could save the map, though, and bring it back later... or even advance it a few hundred years and let a new campaign form on the ruins of the old one.
MMORPG: If a world returns, or one nearly identical to a deceased world, how or -will- that be explained?
Todd Coleman: It doesn't really fit the narrative, honestly. So enough players asked us to bring back a particular map, we would probably modify it a bit, and say it was a world that just happened to be "very similar".
MMORPG: How big will each world be, and how many players will they be able to host?
Todd Coleman: A single "zone" (or continent) is 20 cells by 20 cells, and each cell is 256m by 256m... which means that a zone can be up to 5120 meters by 5120 meters. Zones are laid out on a grid, though, and we can extend the world indefinitely in any direction by connecting more zones in any cardinal direction. Our goal for concurrency in a single world is around 2,000 players.
MMORPG: If too many players want to play on a given world with that particular layout and ruleset, will a second iteration be created or...?
Todd Coleman: Yes, that's the great thing about this architecture, we can bring up as many worlds as we need to support the size of the community. That said, we'd probably change the map some (even if it had a similar layout and scarcity of resources), just to keep the campaign fresh.
MMORPG: Will there be any commonalities between game worlds / maps?
Todd Coleman: I'm sure there will be, but to be honest, we are in uncharted waters, here! What we are doing is very different than "other MMOs", so it going to take some trial and error to get it right. That is part of what makes this whole process so exciting: we (meaning ACE and our players) are going to try it and find out together what works, and what doesn't. We'll try some ideas, learn what works and what doesn't, adjust our plan and try again.
MMORPG: One of the best things about handmade maps is that you get identifiable landmarks and get to know a world. I worry that a world that's random will just seem generic. How do you address that?
Todd Coleman: Remember, they aren't random -- they're assembled out of hand-made pieces. And the history simulator will build on that -- creating the ruins of ancient cities and fallen civilizations. And on top of that, the players will come in and really shape the world.
If you think about the real world as an example, while a lot of our memorable locations are natural formations (like the Grand Canyon or Mount Everest), even more of them are man-made (the pyramids, the Eiffel Tower, etc.) Hopefully this approach will create worlds that are interesting to explore, but also allow our players to make a dent in the universe.