Level With Me is a series of interviews with game developers about their games, work process, and design philosophy. At the end of each interview, they design part of a small first person game. You can play this game at the very end of the series.
Andrew Weldon started as a Half-Life 1 modder and made the notorious ns_eclipse and ns_veil maps for Natural Selection, then worked at Raven on Quake 4, Wolfenstein, and Singularity, then worked at Gearbox on Borderlands, then helped prototype some of the first test levels for Natural Selection 2, then (inhale) worked at Lightbox on Starhawk — and now he’s at Bungie, working on Destiny.
Robert Yang: Thank you for being the token AAA person in this series.
Andrew Weldon: [laughs] That’s a good way to put it.
RY: I had (surprising?) difficulty securing AAA people for this. I understand it, but it’s also a shame that AAA can’t open up more.
AW: I can get why that stuff happens, though, having worked at several places. Anything Phil Fish tweets becomes a news article…
RY: But he’s not “AAA.”
AW: No, but… my friend Manveer [Heir] at BioWare, he had similar stuff happen, with his statements becoming headlines. Fortunately everything turned out okay, but… I’m glad that I can sit here and do this.
RY: Yes, much gratitude to your Bungie overlords.
AW: I think you’re seeing a lot of cool stuff happen on Twitter, with individual AAA developers having more of a presence. We’re interacting more with communities. When I worked on Starhawk, some fans even drove to our studio with their kids. We gave them a tour and I showed them some levels I was working on.
RY: That would intimidate me. A father and son relying on me for their bonding time? If the game sucks, I’d be responsible for their relationship deteriorating…
AW: I’m glad I didn’t get nervous. I have a 7 year old kid myself, and I’m excited to start sharing some of this with him.
RY: Does he want to become a game developer?
AW: Not really. One day he’ll want to be a police officer, the next day he’ll want to be Spiderman.
AW: But, my parents didn’t really know what games were…
RY: Neither did mine.
AW: That’s something our generation has gone through. “You can’t make money doing that. Go get a real job!” Being able to sit down and play Lego Harry Potter with my son, though — I’m most excited about the day he’ll be able to play something that I’ve worked on.
RY: But your work is too violent and graphic for now?
AW: A little bit. My wife plays games too, so we’re good about deciding what’s appropriate for him. Lately he’s gotten into Little Big Planet, and it’s cool to see him place stuff in the editor and come up with his own thing. I’m looking forward to this thing we’ll be able to share.
RY: I like how you make your job so personal, and I think that comes through in your work. Like, when I was playing Quake 4, some parts felt really Andrew Weldon-y.
AW: Think about an Adam Foster level, how he builds in a certain way. He does stuff that I would never think to do. As I’ve worked more in the industry, I’ve become more cognizant of how we all have different backgrounds and we all learn from each other. Like, do you remember how I was building a Team Fortress 2 map with a roundhouse [for trains]?
RY: Yeah, with all those rotated instances and tracks, magically lining up.
AW: That trick came from a co-worker at Gearbox, working in Unreal. Frankly, a lot of my work is mostly just sci-fi themed levels with 45 / 90 degree angles, like the layout for ns_eclipse. But when I see someone else build levels — “oh, you can make curvy things?” — that cross-pollination is important. So on Quake 4, we all kind of worked on everything, honestly. How far did you play in Quake 4?
RY: I played through the first chunk, just after the vehicle section when you get off the ship.
AW: In that first chunk, I worked mostly on the Hangar levels. I was fairly new on the project, and we had just cut a bunch of levels I was working on (but that’s how it is) — so I took over for another designer (who got sick and had to leave) and inherited his levels. They were kind of in the pink-box stage, like how you might orange-box in Source or gray-box in CryEngine to prototype a rough version of a level. I ended up taking those levels through ship[ping]. I wanted to preserve a lot of the original layout…
RY: Wait. Why preserve the layout?
AW: I thought it was a good one. The original designer had an architecture background, so he had this uncanny sense of connectivity for everything. He even modeled out some pipes to loop throughout the entire level. Well, Quake 4 is usually pretty linear, but there’s a part here where you have to defuse some bombs, and you can defuse them in either order. So we spawn some “gunner” NPCs to lay down suppressing fire for a melee “berserker” NPC below, then we also script them with “tethers” that told the gunners to stay within a certain radius around a point. I had to script the positioning so that it worked whether you defused one bomb before the other — and then when the player steps into a certain trigger, it changes the tethers for the gunners to keep the combat interesting.
RY: About the pipes: I noticed them because that’s like something I might think to design. But if I were a typical player, going through the game room-by-room, I’m not sure if I’d notice how it’s the same pipes, or get that sense of connectivity.
AW: Well, the layout did change a lot, that’s what happens when you make games. The first thing doesn’t always work out. Maybe it was too closed, too much of an interior space. If it’s more of an open exterior, you get more of a sense of place because you can see more of it all at once, and understand how each thing fits together. But like… really, this is all just totally made up. Some guy just sat down at a computer and moved some polygons around and thought, “this is good.” These touches or whatever we can do, help ground the space, to the people who do notice them.
RY: If you’ll indulge me, and let me explain your own style to you — I think you have a good sense of “mid-level” forms and shapes. You don’t care about small details like wires, or get lost in huge structures. Your construction feels nice and “chunky” in a way that AAA level design often doesn’t anymore.
AW: “Chunky” is a good word for it. But again, I can’t take credit for everything, some of it might’ve been in the original layout.
RY: I wanted to ask you about this alcove, after you blast through the glass wall. It’s quite detailed, but seems functionally useless. Was there something planned to be here? I smell a story…
AW: I can’t remember off the top of my head who built this, but now I’m bothered by the texture misalignment there, how the panel edges and seams don’t line-up…
RY: [laughs] You mean that upper curve? I mean, I guess —
AW: That gray texture was a “make your own computer console set” kind of texture, where all the different parts and things were on one texture sheet. You can see more of that texture on those supports at the ends of the curves. That lower curve portion is fairly frequent throughout the game. It’s actually made of two separate models — two separate patch mesh strips, to make it easier to texture. And in these alcoves, we’d usually jam in all these monitors to make it look like a control station… but it looks like we forgot to put the monitors here.
RY: So one person made some pieces, and everyone just copied that chunk and put it in all their maps?
AW: Yeah. We didn’t have many artists making meshes for us; we built, detailed, and lit all our own maps. You’ll see a coil of red wire throughout all the levels… that’s a patch mesh cylinder we wrapped around itself. So maybe one designer made that, then someone else went into that map and grabbed it, uncoiled it a little, and made it re-textured it to a different colors. Those things happened a lot.
RY: Do you ever miss having more control over how your levels look? I assume these days at Bungie, or even in your time at Lightbox, you were mostly gray-boxing stuff and building rough layouts.
AW: Working at Lightbox was interesting. They didn’t have a level editor at first, they just had environment artists build the levels and then scripters go in and add the gameplay. But I was greedy, I wanted to be hands-on, so I learned World Machine and a bit of Maya so I could sculpt and generate the terrain directly. Eventually, our level editor became this Lua-based thing that ran inside the game… I used to make fun of those Westwood College TV ads, where they’re developing games with controllers, but then I found myself making levels with a PS3 controller.
I like what they did with the Spark editor for Natural Selection 2. It’s like a polygon-based SketchUp, with very fine control over everything. I’d love for more editors like that to takeoff because they break down barriers for players to learn level design without pirating a huge modeling package that, really, is not suited for level design. I just want to make a box with a texture on it, and not have to manage texture projections on it. When I was a kid, it was great to learn all that pretty quickly.
RY: I miss it too, but for kids today, that’s what Minecraft is? Some New York City schools even have Minecraft classes. A 3D pixel is so much easier to use than a 3D polygon. Maybe we don’t get it, and we need new tools.
AW: I really admire the people who make their own tools to mod the games they aren’t even supposed to be modding. How do you even do that? These days, with so many free multiplatform game tools like UDK or Unity…
RY: That’s what frustrated me about being a modder, your games are held hostage! My friends’d want to play, and I’d say, “download Steam, then download and install Episode Two, then run it once and quit, then unzip to the SourceMods folder inside your…” — to me, that was outweighing the benefit of modding.
AW: But there’s still so much value for full games and mods existing side by side. You’re kind of making this fan fiction thing, about this system and world that you care about. It also lets you focus on really specific aspects of games, to deconstruct just a few pieces and change them. A lot of my first Half-Life maps were about practicing a level design style from a tutorial. They were kind of like “studies” of that teal lab wall style from the beginning of Half-Life. I remember even replicating those cylindrical probe-like things in the test chamber…
RY: Oh yeah, the [anti-mass spectrometer emitters] that spin around. I think they had a barrel texture on them? They were flying laser barrels.
AW: That’s another thing that I miss, putting a barrel texture on a thing that isn’t a barrel, or a wall texture on a ceiling.
RY: And this is how we got into games! We’re romantics. We learned to have emotions with games first, and then learned technique second. I see a lot of kids in game schools now who just learn techniques, and think that’s all that game development involves.
AW: There’s just fewer opportunities for them to learn by modding. A lot of studios don’t release toolsets anymore, and they say that “it’s too complex to release.” That doesn’t necessarily mean, “modders aren’t smart enough to figure it out.” It means that they have this super-specialized toolset that’s hooked up to server farm that’s doing overnight builds of the game, running on really high-powered hardware. It’s a lot of effort to make that usable for regular people. I admire Valve’s commitment to this. In fact, I remember the first time I heard of Valve was when I was using the Worldcraft 1.6 level editor for Quake…
RY: Oh yeah, Valve bought Worldcraft.
AW: Back then, the Worldcraft developers promised editor support for all these other games, and then they said, “oh, we got bought by this ‘valve’ company, so we’re not going to do that anymore.” Oh, come on! “You bastards!”… But then I found out about Half-Life 1 and forgave them. It helps that they haven’t changed the workflow much, that I can always go back to Source and fiddle with maps. Like I’m back home.
RY: Quake-lineage engines always feel like “home” to me. Then I play an Unreal game and, “ugh, this is so unpleasant.”
AW: The iD movement code always feels the best to me. You hit a button and an impulse is applied in that direction and that’s that. There’s no body awareness wrapped around your input, no barriers, no artificial things happening to the camera. Of course, I’m totally biased, I grew up with iD games. If you talk to people who started with Unreal, it’s totally flipped around. But I’d like to think it’s not a coincidence, how Counter-Strike, Half-Life, and Call of Duty, use the same core movement code.
RY: I don’t think it’s a coincidence. I think we’re definitely right. Zero bias.
AW: [laughs] I’ll go with that. I like the engine quirks that turned into skills, like rocket-jumping or strafe-jumping. Quake 4 was interesting because it was a singleplayer sequel to Quake 2 but we also wanted to make a multiplayer successor to Quake 3…
RY: Yeah, the mixed lineage was really interesting to me.
AW: So when we got into multiplayer, we knew we had to preserve strafe-jumping and player speeds. But Quake 4 was built on top of Doom 3 and their base movement, and that is a slower game. And actually, in Quake 2, you could ramp-jump.
RY: Oh, what’s ramp jumping?
AW: When you run up to the top of a ramp and jump, you can sort of maintain your momentum. We actually built that into a few of our multiplayer maps. One map had curved pipes with angled clip brushes, so you could ramp jump up the pipes to get to a higher ledge.
RY: Did players notice? Did they like it?
AW: I hope they did. Maybe not the shortcuts, but they definitely noticed the regular ramp jumping. The thing is that I’m actually terrible at all these things, I can’t strafe-jump for the life of me. But it’s important that it was there.
RY: How do you design levels that support things that you, the designer, cannot do?
AW: I cry a lot. [laughs] Other people did the measurement work for me, like how far you can jump when you have full momentum — and sometimes you just build something, watch people play it and discover new tricks that even you didn’t know about, and then tweak the map to support those better.
RY: I was always too scared to make multiplayer maps. I feel like you have to be a high-level player, to have enough understanding and skill, to make good multiplayer maps.
AW: It doesn’t hurt. A lot of my co-workers can definitely play at that level. But how do people design golf courses? Those guys probably aren’t Tiger Woods. And with baseball diamonds, every Major League outfield has different measurements.
RY: Was that on purpose? They didn’t think, “let’s put the Green Monster here to do __” did they?
AW: In some of those stadiums, they have custom rules built around things like that. I was a big Houston Astros fan, and the Astrodome had these hanging speakers —
RY: You realize we’re talking about baseball on a British PC gaming website?
AW: [strange retching groan noise?]
RY: Wait, no, okay, keep going! Sorry. I’m interested.
AW: — the speakers. Players would start hitting balls off them. But how do you handle that? That’s not in the rules of baseball. They ended up ruling those as ground-rule doubles. (Which is when the baseball bounces off the ground, and over the fence. It doesn’t count as a home run.) It’s not a star player designing these things, you just need to understand what the stars are capable of.
RY: It’s so hard to try to put this process into words though, right? This understanding? Why do we make things the way we do? Like, how did you come up with the layout for ns_eclipse? You just wing it?
AW: Usually. But for ns_eclipse, I decided I would take a bunch of graph paper and draw the entire layout, to scale, across all these pages. I thought I was brilliant, for sketching this perfect thing and knowing everything!… And that was such a naive thing to think. To my credit, two areas stayed how they were, but I rebuilt that map at least six times. Plus, as I was building it, Natural Selection was changing in all these ways too.
RY: That describes so much of game development, right? You’re trying to hit a moving target.
AW: Exactly. You don’t have the luxury of making a game and then making all the levels after. Everything is changing constantly. When we got to the end of Quake 4, we all looked each other, and thought — now we know how this works, now we know what we were making, we can do everything better in half the time. But we couldn’t, because we were done and we shipped. But that approach isn’t great either. With ns_eclipse, I kept gutting 75% of the map after a playtest, and that’s just as bad as overly planning everything.
Now, I think it’s best to start with a core idea. It will probably fail, but you will fail in the right direction. For the Quake 4 multiplayer map pack, I made a level with a rocket launcher at the top of a large room in the middle of the map, and all the side areas snake around it and feed players back into the main area, like a swarm. It ended up being a terrible and unbalanced idea, and I replaced the rocket launcher with some armor instead, but the original idea with making these opportunities for cool line-of-sight moments, carried though.
RY: Those are my favorite maps, the ones that are “about” something — about a tower in the middle, or a jump pad path in the middle. A concept that holds it together.
AW: It’s the “personality” of a level. A rail gun alley isn’t just based on the level geometry. It’s also based on the energy, the potential for cool things to happen. There’s a joke about how much time level designers spend, flying the level editor camera through a level without making any changes…
RY: But it’s time well-spent. It’s testing the feel of the space. I used to think it was a waste of time, but I’ve changed my mind.
AW: Yeah. It’s the best thing you can do. But it’s funny when someone’s watching you, it looks like you’re doing nothing, just panning around, staring at it. Then you stop for a moment, select one thing, and nudge it over 1 inch. Over and over. But I think we’re actually inventing these player experiences in our head, thinking about how people will navigate and look around. It’s very deliberate, but to most people, we look like idiots.
RY: But we’re not idiots! We’re not!
AW: These levels are playgrounds, and playgrounds are very interesting to look at and move through. They’re also good at making you feel awesome. I think that’s what we do: we help players be awesome.
RY: Okay, so… this first person prototype is rough, please have mercy, etc. Right now, you just pull yourself along with these gray poles, and then go from there.
AW: [playing] I had a hard time figuring out why I was going at a slow pace, and how to move in range of the pylons. It reminded me of the grappling hook in Quake 2 Capture the Flag, but I imagine that wasn’t your inspiration? The momentum is interesting, with grabbing one and swinging yourself to the next one. I could imagine this being a skill-based gravity hook thing.
RY: We can turn it into that, if you want.
AW: Nah, I don’t think it needs that. I’m more interested in that thing that looks like a light pole? I saw the arm reaching across. I thought I’d be able to swing from that arm. What if I grab that and throw myself a bit further, slingshot style?
RY: Okay, I can add that. I’ll add a momentum hack property, just for that arm. And I think that’s an interaction that makes sense at a lake, to swing yourself into the water.
AW: Like a tire swing. Yeah. What if you raised that area around the water? So that you can take a sort of “leap of faith” and swing off from a height? But I don’t even know what you’re intending here. What am I? Why am I moving this way? That’s fine if that’s left unanswered…
RY: You can answer it if you want.
AW: I don’t think I have an answer. I’m more interested in how you see the ripples in the water, or how the trees frame your view of the moon up there. My main concern is if all of that gets raised up so you can swing off of it, then this view won’t have the same effect.
RY: A lot of this work was me, interpreting the person who came before you. She talked about a green lake and seeing the trees and such, and I tried to execute that view in the spirit of what she said.
AW: It’s really interesting to see how she was more interested in painting a picture of this scene (and it works) and then I come in and start thinking about these Quake 2 movement mechanics…
RY: [laughs] Maybe that’s what this needs?
AW: So, if we raised a cliff… you could push the water further back and see more of the water. Then suddenly you’d come up to the edge and notice it’s a big drop-off. The other thing I’d want is that there’s something in the water, but you don’t see it until you’re in there already. I’d want to get a hint of what’s down there, before I take my leap of faith and jump-in.
RY: Anything else?
AW: No, I think this will change things enough for the next round.
RY: A very reasonable addition… very well scoped. How AAA of you.
AW: Funny thing is, I’m usually terrible at this sort of thing.
RY: Thanks for your time.