Making Epic 3D Dungeons (Part 1)

Posted by Rampant Coyote on May 11, 2010

UPDATE: Part 2 of this series, dealing with practical considerations for level design, can be found here:

Making Epic 3D Dungeons, Part 2 – Applying the Lessons

I grew up playing D&D, sometimes using the classic modules from TSR, often penned by Gary Gygax himself, or other writers like Allen Hammack, Ken Rolston (known to modern CRPG players as the lead designer for The Elder Scrolls III: Morrowind and The Elder Scrolls IV: Oblivion), and Lawrence Schick. The modules had these really neat maps in light blue ink. The light blue ink made them difficult to copy by many of the copy machines of the era, so that was copy protection circa 1983.

As a result, in my mind, that’s what a “dungeon” (to use the generic term from fantasy RPGs) looks like. Something like this. Or this. Or the one on the right.

In 1988, I attended a lecture by Tracy Hickman, author of a lot of … well, juvenile fantasy literature back in a time when such a thing hardly existed. He was also the author of the classic D&D module, Ravenloft. The lecture was specifically about adventure creation, and he spoke a few minutes about maps. He commented on how, if you tried to create a 3D model of one of the castles from these classic D&D modules, it would come out very “dumpy” and un-castle-like. That was part of his thinking when he designed the castle in Ravenloft. He tried to make it more realistic. He also discovered how terribly the third dimension tripped up pencil-and-paper gamers. They’d be convinced that some awful teleport trap had happened, when in reality they’d just made a mapping error.

The early first-person-perspective Computer RPGs worked well translating the old 2D graph-paper maps. These games offered four directional views in a tile-based world. They really couldn’t show curves or any walls oriented in anything but the four cardinal directions, but they got the job done okay.

And of course, the top-down or isometric view CRPGs worked fine with these kinds of maps, although again everything worked best at nice 90 degree angles without very many exceptions. The descendants of these latter games still stick to the nice 2D, rectilinear maps pretty well, though they often offer more interesting wall and floor shapes and height changes.

But Ultima Underworld was the game that started screwing everything up.

For the most part, it stuck with safe 90-degree angles and so forth too, but it added the vertical element to its maps. And underground rivers. And occasionally broke out of the rectilinear world to give us 45-degree corridors. Our worlds and perspective would be forever changed.

Yeah, Doom didn’t help much, either. But it came a little later. Ultima Underworld was ahead of its time in many ways. But still – those maps bore enough resemblance to the old 2D maps of the classic Advanced Dungeons & Dragons era that I was in familiar territory.

So anyway, fast forward to today. Making 3D maps for a first-person perspective game. Tracy Hickman was 100% correct. What I think of as an awesome, classic, “old-school” map – like one of the above examples – comes out awfully , squat, boring, and – yes, dumpy – if translated directly into full-on 3D. Oh, and they also turn out pretty maze-like, too. And about 2/3rds of their actual size relative to what should be your “proper” eyepoint. I’m sure there are some things that can be done to make it more interesting and visually appealing (Kevin did wonders for the Temple of Pokmor Xang, which began life as a 2D graph-paper map with very few elevation-change cues). But apparently the direct translation of my vision of a “good map” runs counter to that of visually appealing and functional 3D map design.

And to make matters even more complicated, the needs of an RPG run counter to the principles of good FPS map design, too. As Brian, my another map-maker for Frayed Knights (and another family member roped into helping on my insane projects) has discovered. In an FPS, particularly competitive multiplayer levels, very circular map design is called for, with lots of “choke points” where paths converge on each other. You want very few if any “dead ends.” On the contrary, in an RPG, exploration is the goal, and finding all those end-nodes and discovering what surprises they hold is the whole point. He’s still fighting a few habits on that.

Meanwhile, I keep learning that I’ve got to use the vertical. vertical = interesting. I took it to an extreme recently and built a “mini-dungeon” constructed entirely in a huge vertical shaft.

Anyway, it’s not like this is totally uncharted territory. There are plenty of books (I own one) and articles out there on level design. It’s not like there was anything sacred about those old-school Gygaxian maps even back in the day. But there’s one thing to understand what’s right and what’s wrong, and what should be done when making a 3D RPG map, and something else entirely to build the thing so that it neither looks nor plays shoddy.

On my end, there’s a lot of learning I still have to do. And a lot of habits to break. And a lot of skills I still have to develop. These maps take a lot of time and effort to create.

I do miss the old 2D maps sometimes, and get just a little jealous of those indie RPG developers who still get to work in the somewhat more flat, tile-based realm. There’s a lot to be said – and a lot more that can be done – within that simpler framework. But it’s also very fun (if a lot of work) to embrace the third dimension, learn to love the vertical, and do the kinds of things that Tracy Hickman could only dream (and write) of in 1988.

Filed Under: Art, Design - Comments: 26 Comments to Read

Justin Alexander said,

For some old school tabletop inspiration in 3D mapping check out Paul Jaquay’s Cavern of Thracia (which was published by Judges Guild).

For more useful lessons from the tabletop, I also recommend checking out this post at ENWorld. It isn’t specifically discussing 3D design, but it does have some incredibly useful insight into what makes for an effective flow in a dungeon design.
Rampant Coyote said,

Yup, I loved Melan’s article – it’s a classic (I linked to it a few weeks ago.) I was very impressed with how much effort he put into that analysis.

Caverns of Thracia – I looked at the original version of that module several years ago, but don’t have a copy. I was tempted to pick up the Necromancer Games remake (now only $18 in PDF format). Not sure where I can find the original – doesn’t look like DTRPG or Paizo has it.
Bad Sector said,

The FPS level design principles you mentioned are true for multiplayer maps. Single player maps, while not exactly the same, have more similar principles to what you are looking for.

You may want to check some Quake 1 maps at Quaddicted (filter the map list to show maps released the last 4-5 years). The people who make Q1 maps have brushes for bread and butter and Q1’s abstract design has a very dungeon-y look to it. Although the game’s core is still a FPS, there are a lot of ideas explored in some of the maps both from a visual design and a gameplay perspective, thanks to the game’s abstract and open nature.
Flux said,

While I do appreciate the greater possibilities and more organic feel of a 3D environment, my main problem with it is one of speed. The old school, first-person, tile-based maps effectively let you move as quickly or as slowly as you wanted. With movement in 3D, you’re locked into whatever speed the developer decided on. Take Wizardry 8 – its the first (and only, at least in the US) game in the series in full 3D, and I did enjoy it a lot. However, there were times, particularly when I had to backtrack over an already explored area, where I wished the series had kept to the original style. Holding down the ‘move forward’ key for several minutes at a time with nothing else to do didn’t exactly enthrall me; its just busywork.
Rampant Coyote said,

Flux – you aren’t wrong. But I think that can be mitigated by game design. Look at the Baldur’s Gate series (which is 2D but had the same issue) – once you’ve been to an area, it’s permanently part of an auto-travel map. Many more modern games (including FK) have adopted a similar system to reduce any extra back-tracking.
Bargeral said,

One of the complaints I have about almost every 3d game I’ve seen is how the scale is just off. I mean when was the last time you’ve seen a ten foot door in real life? The real world is much smaller then it gets represented in 3d. I’d love to walk through a scale model of a real space. I suspect a lot of the scale is oversize to avoid issues with getting stuck on the geometry and such. A simple wooden chair that you might have in you kitchen doesn’t have legs made from 8″ dowels, nor come up to your waist. In defense of large doors, real life doesn’t ever have to accommodate trolls, but still why would a hut lived in by a farmer have one that large?
Rampant Coyote said,

Awesome point, Barg! I’m glad you brought that up, because this was something that surprised me, too.

The part about collision is definitely true – we’ve had to make some changes to the castle from some of the real-life floorplans we’re using as a template simply because in real life, people can twist, turn, bend, and duck really naturally and almost automatically. Not so much in a game. And those castles were very specifically designed NOT to be easy to navigate.

But the other point is more subtle.

When I first started trying to do 3D environments (for a game I’ve since back-burnered and currently plan to release as more of an iso-view game once FK and AC are done), I thought exactly the same thing, and I started modeling everything exactly to scale for somebody with a 5’6″ viewpoint. I thought it would be awesome…

And it was terrible! It felt like I was Gandlaf wandering around Bilbo’s house in The Fellowship of the Ring. Everything seemed like it was shrunk by about 25%. I heard Brad Bird make a similar complaint in his commentary on making the movie, The Incredibles, so it’s not just interactivity and first-person perspective. But that viewport shrinks things.

I thought – and still wonder about this – if it doesn’t have something to do with how we look at the world. Are our eyes generally focused on the where our feet are going to go? Do we tend to look more down than up? And is the reason things look too small in the game world because the eyepoint is looking straight ahead at an unnaturally high angle? Maybe. But when I tried to force the issue in Frayed Knights with a fixed eye pitch, people REALLY didn’t like it, and I don’t think it helped much, either.

So anyway, after these experiments, I kinda threw in the towel and did what every other game does. But it wasn’t for lack of trying!
Calibrator said,

Apropos Melan:
He has made some incredible fan missions for “Thief 2”!
His city-style missions perfectly capture the atmosphere of the game setting and are expertly built.
Rampant Coyote said,

You know, I remember him talking about that, and I never checked out his mods. Do you have a link?
Gravity Denier said,

I read that entire article looking for useful tips. I think the conclusion was ‘build it so it looks and plays good.’

Insightful.

And you disagreed with dead ends. Beg to differ. Elder Scrolls Oblivion got that right. While the rule should not be mandatory, there is nothing worse than fighting your way 7 levels down in a dungeon, then turning around and painfully plodding out the way you came. A good design should lead to an exit near the end. It shouldn’t be used in all cases, else it’s predictable and immersion breaking, but certainly most.

Dead ends work only in tabletop, “Ok then, we’ll head back to the junction-…” *slides paladin miniature*

In a computer game it is painful. A pain that escalates exponentially with every successive dead end. There is no tension created here, no atmosphere, no wonder or mystery for this unexplored dwarven catacomb. Just a hundred cubed feet of suck.
Rampant Coyote said,

Okay – I guess I should have rephrased, then. By dead ends, I’m thinking “Leaf Nodes” in a graph of the dungeon. A room. A hallway with three rooms in it that goes nowhere else. Yes, backtracking may be involved, but it’s not necessarily going back through the entire dungeon. Every room with no exits is a dead-end, by my way of thinking. But I also view each room as a goal in and of itself.

In a competitive FPS, that’s generally a no-no without a really good reason. That’s a death trap. The idea is to always have people moving, and generally allow a point of egress that’s never too far away.

The usual solution for a largish dungeon (for us and others) is to have some kind of a one-way door that only works from further in, a teleport system, something that can only be unlocked from the other side, or a drop from gravity with no “fly” spell to shortcut your way back up again.

So I agree with your point about backtracking, but only if it’s excessive. That’s boring. My cardinal rule of gaming (which I hope not to violate too much) is “don’t be boring.”

As far as more specific tips than what I offered here (which is more “what to avoid” than “what to do”) – I’m still hunting for more. Another point, for larger dungeons, is to make the end-goal visible earlier on (I haven’t followed that always, but it’s good to have that in the checklist), and try to allow multiple ways of navigating the dungeon, or at least what order you are doing things, and have consequences for your choices. These may not be good / bad consequences – just differences in how you have to deal with things.
Rampant Coyote said,

Heh – I may have to create a “Making Epic 3D Dungeons Part 2” 😉
Xenovore said,

In competitive multi-player FPS games, the two key reasons for circular level design are 1) to maintain access to resources, and 2) to maintain access to pertinent areas. In other words, all players need access to weapon upgrades, ammo, power-ups, etc.

For single-player (or co-op) games, particularly RPGs emphasizing exploration, the main reason for using circular level design would be to reduce or eliminate redundant travel, i.e. back-tracking. But there is another reason which may potentially be more important.

Circular level design provides the player(s) with choices, i.e. multiple routes and possibly multiple solutions. (Deus Ex is a fine example of this.) Also, the players may choose to back-track to fully explore the alternate routes, but it’s their choice to do so, so there’s no feeling that they were “railroaded” into back-tracking.

With that said, dead-ended routes (i.e. with forced back-tracking) are certainly applicable, but need to be kept short and/or interesting — keep the player engaged and having fun. For example, you could inject some new game-play elements into the back-track. L4D2 does this on one level (“Hard Rain”), where the entire second half of the level involves back-tracking, but under considerably different conditions. This keeps it interesting, even though it’s an area that the players have already seen.
Gravity Denier said,

Oh, Leaf Nodes-… or, rooooms.

Wouldn’t be much of a dungeon if you couldn’t have rooms, Coyote. I think dungeons comprised of just hallways are called Tunnels.

One Tunnel. No deviation. And up ahead, a Gelatinous Cube.

Since the article is about 3D dungeons you should discuss more puzzle navigation (which also fits with what you said above about visible end goals). 3D maps allow you to see the golden object of your quest tantalizingly out of reach on a walkway above. Find out which lower level path leads up there should be the puzzle.

And personally, I love a nice big pit. Open and ominous. While you’re fearfully peering into it wondering whether to descend into it, or brace yourself for dragonflame the player won’t be looking behind them. Which is useful to move up a party of slowly moving zombies.

I love a red herring.
Geoff Dunbar said,

Great post.

A lot of the pain of 3D is felt by the programmer as well. Your pathfinding, AI, heck, even line of sight, become much more complicated. That stuff is a lot easier to deal with in good-old-fashioned 2D.

Geoff
Xenovore said,

I should have noted this above… “Circular” implies that the player will be returning to areas previously visited, but in a non-linear fashion. This may ultimately be just as egregious as linear, back-tracking level design, if those areas are being constantly re-visited.

So, let’s call it “multi-route level design” which more accurately describes the best way to design levels, IMO.
Bargeral said,

Leaf nodes. So one thing that irks me is when I get a choice between two paths and I know one will lead onwards and the other will be a dead end. I want the dead end so I can explore the entire map as efficiently as possible. When I accidentally pick the “right” way I have to back track, but it’s never clear if the way I am going is going on or just about to end. Maybe it’s just a very long dead end? Or maybe I should go back? Or or…. It was much easier with square maps and without seamless worlds. It’s the problem with the underground places that go on and on, with a hut you can tell really easy how many nooks to poke your nose in, but underground it could be anything. I like big doors that scream, this way to advance the plot. I can back slowly away and go “finish” the area first. That and huts. I like huts.
sascha said,

Great article! I love reading about RPG dungeon mapping.
Rampant Coyote said,

Well, please feel free to contribute your $0.02. We’re all still learning here.
Maklak said,

I eventually got used to it, but one thing that struck me in many games was how impractical the buildings seemed. I guess I’d opt for a more “realistic” design, where and when it wouldn’t interfere with the fun too much. Multiple entrances/exits (some may be hidden, or unaccessible when you start to explore) apart from main entrance. Large buildings modelled after real buildings. Add kitchens, toilets, bedrooms, and so on. I gnerally like things to make sense, not just follow the rule of “cool”.

If there are monsters, they should be “themed”, and not randomly thrown together, like (for DnD) Orks living next door with owlbear, gelatinous cube, and some undead, each patiently waiting in its own room to be killed by adventurers while munching “giant rats”. DnD modules did a reasonably good job with this.

Monster tactics could involve something other than “fight to the death” and “escape for 5 seconds, then fight to the death”. Defenders of a fortress should escape from adventurers when loosing (1-2 escaping, while the rest slows down the party works even better), team up with reinforcements, and use choke points and traps to TPK. There are two problems, I see with this tough. “Intelligent” monster behaviour is difficult to progarm, and probably frustrating to players. To take it to the extreme include alarms calling in guards/cops/monsters from the whole area.

Automap is a nice feature. Descent had a good one in 3D, it could be rotated, zoomed and all. That made it much easier to explore everything before finishing the level. It was clearly visible, where everything was explored, and where not.

I’d like to see a dungeon where there are few or none combatants, but the environment is schizoid. Something like in Escher pictures, where the geometry is just wrong, gravity changes depending on where you are and stairs move, like in Hogwarts. This would be tough to invent for the designer and maybe frustrating to player, but would be interesting nonetheless. This was done quite well in ending sequence of film “The Labyrinth”, and not so good in “Hellraiser 2”.
Justin Alexander said,

Speaking of designing old school D&D spaces in digital realms: http://www.whitewolfbrigade.com/Stronghold/Stronghold2_KOTB.html
Rampant Coyote said,

Oooh, what game / software is he using?
Rampant Coyote said,

Ah, just figured it out – Stronghold 2.
Calibrator said,

> You know, I remember him talking about that, and I never checked out his mods. Do you have a link?

Sorry about the late reply but here is some of Melan’s work:
http://thiefmissions.com/search.cgi?search=gabor&sort=author#m

“Bad Debts” was a fantastic first effort for Thief 2 while his masterpiece as of yet is “Disorientation” (not only IMHO). It’s perhaps the best fan-built city mission I’ve ever played, up there with “Life of the Party” (Thief 2 original mission) and Purah’s legendary Calendra missions.

“Return to the City” is his first, promising mission for “The Dark Mod” (a Thief-like total conversion for Doom³ mimicking the Thief-games as best as possible but with better graphics and physics).

I can see a job in the industry for him – which wouldn’t be that surprising as at least two Thief-“master builders” got recruited (Anthony Huso (Purah) working for Arkane Studios on Dark Messiah and David Riegel (Sledge) on Thief: Deadly Shadows) – but I would rather see him build a big Dark-Mod-mission…
sascha said,

Sorry for throwing such an irrelevant comment in before. I was short on time. 😉

Personally I enjoy those typical rectangular maps very much, even in a 3D environment. It’s very unrealistic but somehow fun because it’s the same as with simplified 2D maps: it allows your own imagination more freedom. Maybe that’s just me because I’m such an oldschooler and have fond memories about old games but modern game maps, as detailed as they are, take away most of your imagination.
Making Epic 3D Dungeons, Part 2 – Applying the Lessons said,

[…] couple of weeks ago, in Making Epic 3D Dungeons, I talked about the evolution of the classic “dungeon” paradigm from dice-and-paper […]