Robin Walker is a legendary game-industry designer, known for his part in the creation of Team Fortress, the famous modification for Quake that simply rocked the online world. Robin, along with John Cook and several other of his Australian friends, coded up the mod in his spare time, and it rapidly escalated in popularity, overtaking many commercial products. Soon afterward Valve Software acquired the fledgling outfit, and Robin and John moved to the US to work with Valve on Half-Life. The rest, as they say, is history. These days Robin and John are still at Valve, masterminding the massively anticipated Team Fortress 2, as well as some other unannounced games. We caught up with Robin in his hometown of Melbourne, Australia, where he attended the second annual Australian Game Developers Conference last week.
GameSpot: First of all, Robin, let's talk about Team Fortress 2. Are you coding another Machinima.com intro movie, with classical music?
Robin Walker: Oh wow. I don't know, to be honest. That was really just a pet project between myself and Damien and John [Cook]. You know, we're now but a small portion of the team, and it's something I'd love to do, but it's certainly something that took a significant amount of our time, and back then our time was pretty much worthless. Now it's pretty expensive. So I don't know. I'd like to say we will, but it's likely we won't. Maybe it's something I'll do on my weekends with a couple of friends or something. Certainly I think we know what we'd do if we did one. We've done that much thinking about it. We know exactly what it would be if we ever do it.
GS: So is playing the commander mode any fun?
RW: Yes. It is a lot of fun. We actually have a backup plan for that, so we're not too concerned about that. What we're looking at is, our view is that the commander really [takes on the role of] an RTS [real-time strategy] player, which is why we're tentatively calling TF2 a first-person strategy now. There's some kinky buzzwords for you. We really are designing the commander to be something an RTS player can move into very easily. The interface is identical to an RTS's - you can get in there. Everything they're used to works, like being able to hotkey groups of players, drag and select, and shift to do multiple commands. All that sort of stuff works, and we really hope to entice RTS players into trying this sort of game. It's interesting. I've seen a few stories recently where people are saying that not much interesting has really happened to the RTS genre in the last couple of years and that we're still making sort of similar RTS's, new units, and new features, but nothing has really sort of "woken up" RTS players for a while. And in some ways I'm wondering if we might be the first people to do that. We're going to present [players] with a problem where they have some units that follow their orders perfectly, because they're computer controlled. But they have no ability to act on the fly or change according to circumstances. And then they have these other characters that are real players who don't necessarily do exactly what they want but at times are able to react more quickly and far more decisively than the commander can. I think a good commander will be the sort of person who can vie those two types of units off, and from them produce something really interesting.
GS: We're just imagining getting chewed out over the radio by the commander when you do something wrong.
RW: We still don't have any actual way for a commander to actually negatively affect a player, so we don't have electric shock therapy if the commander doesn't like what you're doing. And I don't think we'll need that. I think the commander will still be fun, and I really do believe that players fundamentally like to have stuff to do. And there's an enormous number of players out there right now who happily capture flags or kill other guys. If asked by the commander to do something else, like guard this area until the enemy shows up, while this other guy does something, then it's just another thing they get to do, and if they succeed, they've made another person or an entire team happy.
GS: Audio: Prerecorded, real-time, or text-to-speech?
RW: We do real-time voice over the Net, so you are able to talk to each other. There's a couple of ways we do that. We do it on a proximity basis so people around you hear it, and we do it on a squad basis, so people in your squad hear you. We avoid problems of people talking to everyone at once, and you can't talk to your enemies and stuff like that because we really do want voice to be something that adds to the game, doesn't detract, doesn't make people more nervous about it. We really hope to be able to do stuff like allowing you to deepen your voice and stuff like that. So we want to preserve the sort of anonymity that people have these days. But we definitely want to do voice. There's a side bunch of us who really hope to get text-to-speech in there. We have a really rudimentary version of it now. It would be really good to get that in. I don't think it will be there initially, but it's something we might roll out in an update.
GS: Can you give us an example of a situation where Parametric Animation rules over an older technology?
RW: Sure. The main thing Parametric Animation allows you to do is to show a greater range of states of other players around you. So imagine at anytime in the game, a player has a number of states. So, in an older FPS (first-person shooter game) I'm running along, say I'm strafing to my right, there's an enemy in front, and I'm reloading, or something like that. Now in older technology you were forced to make decisions about what animation states you had to show to the players. So you're over there and you're watching me. And what you see depends on what the game considers important. If it thinks shooting is really important, you'll see me play a firing animation. Maybe they have firing animations that are created with a strafe, so you'll see me strafing and firing. But an artist had to create firing standing, firing strafing left, and firing strafing right animations. All of that. It becomes a lot of work. Then maybe if you wanted to see me reloading, the artist had to create reloading while running, reloading while strafing. Maybe I'm jumping at the same time, so you have to get strafing, jumping, reloading. All of that stuff. Parametric Animation allows our artists to create every one of these things once, and we blend them all together. So if I'm strafing one way, shooting another way, reloading and jumping at the same time, the animations just all blend together. And the end result is we get extremely lifelike animations. So that's one thing you get - increased states. The other thing you get is that bodies move far more nicely. An example is that we don't actually take four animations and blend them. In Half-Life, for example, we did a sort of semi-step up for the animation. We allowed the player's waist to be decoupled from his torso so you could run left and right, and your upper body could do things. With Parametric, we get a step over that which allows us to do whole-body blends. So that when I look up and down it's not my torso swiveling - my whole body shifts. If I turn and I'm running this way, my weight shifts on my legs, and my arms move. The whole body moves as it should. And it's interesting, but it does make the game a lot more appealing to people who don't play FPS's normally. The players move as they should. They find it very strange when they watch a game and they watch a guy running forward, moving right. It seems very strange to them.
GS: Objectives compass - 3D or not?
RW: It's not 3D. We're moving to a much more realistic feel. Our maps are modeled after much more realistic world spaces, and believe it or not, most world spaces are pretty flat.
GS: Not ten layers of tunnels?
RW: No, exactly, that's right! So we looked at doing a 3D compass, and we thought the added complexity of having to show that in a compass and players having to account for that was not worth it when we looked at it. For example, a couple of our maps have real actual areas where two or three people could cross over. And that feeds back into the commander itself, actually. With the commander having a top-down view, we don't really want to have a ten-story building, because players could be in any one of those stories. Because you're looking down from above, it gets really complex. Our maps are traditionally pretty flat.
GS: Can you tell us something about the onscreen advice or helper AI?
RW: Yeah. Basically it's a sort of a helper bot. You can think of it as a bot that's looking at you and figuring out what you're doing and what you understand or don't understand. Basically it looks at things you've done and things you haven't done and tells you if it thinks you should be doing something. For example, say you're out of ammunition, and it's been 30 seconds. It might say, "Did you know you can reload?" That's a bad example because we actually do automatic reloading, but it feeds into things, like if you're a medic and you're running around with the medikit out, and a friendly shows up who's damaged, the helper bot will point out, "This guy over here has taken some damage. You can go over and heal him." We do that with a bunch of simple rules that try and teach players to play the game. This came out of watching a lot of players trying to play our games. And the commander was another facet of the solution to this big problem, which was when we watch players jump into Team Fortress, or any of these sorts of multiplayer games, they don't know what to do. You put them in a deathmatch, it's pretty simple, they grab guns, and they kill everyone. But if you're going to get into any kind of team-play game, or any sort of more complex objective, players jump in and they don't know what they should do. And the commander is one of our ways of solving this. This helper bot's another one. It tells the player what to do. And we found with players that they don't mind being told what to do; they're really happy to do it. It's just they don't know what to do in the first place. So telling them when they join a map when they first start that they should just pick this particular class, run up here and just defend this area, players have a lot of fun doing just a single, simple goal. And from then on, they're more interested, maybe in learning some other stuff, some of the other classes. But they get a lot better first-time experience than when they just jump in, and there's a bunch of people running about like crazy, they've got a gun, and they don't know what to do.
GS: Regarding the new modern military theme, did you look at any modern military roles or structures to help define the individual classes?
RW: Oh absolutely. Pretty much everything within our new classes has been designed that way. We looked at an enormous amount of military stuff. We didn't look at any specific one. We just search on the Net. Usually I'm looking for something specific. You know, I'm looking for a guy with a weapon that's reasonably high powered. I want it pretty inaccurate at certain different ranges. I want it to have a clip of say anywhere between 30 and 60 rounds, and I want it to be pretty lethal. I mean that's the gameplay I want. And usually I can go out there and I can find a gun. I can find something that fits it pretty well, and so we'll use that. If we don't, then we just create it or we modify something. We know the gameplay we want; we try to fit some reality into that. But we're not really that fussed about it. We don't care if our guns don't have the right number of clips or anything like that.
GS: About destructible components of landscape, are there any major upgrades in that area in comparison with Team Fortress Classic?
RW: We do some things like bridges. We have some demolitions missions where you have to blow stuff up. And when you blow stuff up you get to see a big thing get blown up. And we do some really neat physics stuff to make it look really, really cool.
GS: Flying shrapnel, things like that?
RW: Oh yeah, all that sort of stuff.
GS: Are there footsteps, and how much of the game has a sound or stealth component?
RW: Yeah they're in there. We do footsteps. They are according to the environment. I mean we feel that players should feel like they're a part of the environment at all times. Have the environment react to them, to running across hard ground or running through mud - that should be reflected. So we have footsteps.
GS: In line with the realism aspect, would you say that the weapons will be relatively more deadly or ballistic when compared with Team Fortress Classic?
RW: More so than Team Fortress Classic, but we're still not pushing for a "one shot, one kill" environment or anything like that. The fine line we're trying to go for is that players can't be killed in one shot, but they still feel like they can be killed with one shot. Which is strange to say, but we think we've got a way of doing it. So we want players to take advantage of cover. We want players to work together in a team. Stuff like that. But we really hate the inherent sort of randomness you get when you say "I can shoot you once in the head and you die" on an Internet where your ping is up to 400 milliseconds. There's no ability in such a system for you to feel that when I shot you and you died, I was justified in killing you. We're in a system that is not inherently instant or timely enough to allow players to believe that that stuff is capable of happening. So we're towing a fine line. And it's really just player testing that's going to get us there.
GS: Have you incorporated modern grenades or variants?
RW: We've based it on a few grenades. If anything, we've cut down on some of the wackier grenades we had in Team Fortress Classic, which were a lot of fun, but they don't really fit in with the world we're trying to create with Team Fortress 2.
GS: Do you have a smoke grenade?
RW: We have a smoke grenade.
GS: In terms of effects, have you revamped things like ejecting shell casings, or gore?
RW: We've got a bunch of effects. They're pretty damn cool looking, but, well, you'll see. (laughs)
GS: AAs for level editing, will it be a version of WorldCraft, and will it be shipped with Team Fortress 2?
RW: Oh yeah. Absolutely.
GS: And what about the bots. How are they going?
RW: We're scaling down our bots in some ways. I don't really want to talk about bots too much, because it's still something we consider too much. With research. I think it's something we'll talk about pretty soon.
GS: Considering every version of community-created Team Fortress bots that we've seen, it seems like it must be almost impossible to make them behave correctly given their different classes and various tasks they would have to undertake, in a range of different situations. Even the most recent community-made Team Fortress Classic bots still drown, still do foolish things.
RW: Bots are an incredible problem. Well one of the approaches we're taking now is trying to have simple behavior through the bots, complemented through the AI of the player who is the commander. So hopefully the commander will be giving the bots the higher-level intelligence, and we can cut down on trying to have an AI that understands everything about the game.
GS: Can we please have the game by 2005?
RW: Ah. You'll absolutely have it way before then. We've been working on a lot of stuff this year, not just Team Fortress 2.
GS: Like Powerplay?
RW: More like stuff with Team Fortress Classic. I mean we're continuing to roll out updates for Team Fortress Classic that have a lot of content. We're doing a lot of mod support. We're supporting a lot of the guys like Counterstrike and Gunman and trying to make sure that all the mods out there that haven't hit it big have everything they need. Been working a lot with mods like Frontline Force and Wanted. There are a lot of mods out there that are doing really good stuff, and we want to make sure we support them. Our focus for the last year has really been on trying to grow our numbers online. An interesting thing I think we decided was that the number of players playing all our games at the time we ship Team Fortress 2 is as important a requirement for Team Fortress 2's success as, say, having Team Fortress 2 work on different accelerator cards. If we ship Team Fortress 2 at a time when there's like 500 people playing our game online, then Team Fortress 2 has no chance of success. So we consider having an existing community at the time we ship Team Fortress 2 to be an extremely important requirement for Team Fortress 2's success. So we've been spending a lot of time growing our online user base. I heard we hit about 32,000 simultaneous players four days after we released our patch last week, which I think is our new record. I mean that grew from something like 5000 simultaneous players about a year and a half ago. So most people probably don't realize exactly how much time we've been spending, but we've been spending an enormous amount of time growing those numbers online. That will all pay off when we ship Team Fortress 2, and the day it ships there'll be thousands of players out there just waiting to get onto it. Multiplayer games are only fun if there are other players.
GS: What can you tell us about the in-game atmosphere you are creating? How are you combining the atmosphere of, say, Aliens or Saving Private Ryan into it?
RW: Well, we're not really trying for Aliens. And to be honest, we're not really trying for Saving Private Ryan too much. I mean, Saving Private Ryan is a goal in some places, but not a goal in others. Saving Private Ryan as a movie was trying to show you how horrible war was. If you watch the movie and look at all the characters in Saving Private Ryan, they're not happy. None of them are bright, none of them are clean - they're all in war. So to some extent we want to create the environments of Saving Private Ryan, perhaps, but we certainly don't want to create the atmosphere exactly as Saving Private Ryan, and we don't want the characters to feel like they're in Saving Private Ryan too much. Because Saving Private Ryan is a horrible world to be in. If you wanted to look at movies we thought were more like it, I think internally we've discussed our atmospheric goal as something somewhere between Saving Private Ryan and Calley's Heroes. I mean Calley's Heroes is this film where all the heroes are supermen. They're all these incredible guys, but somewhere between it, we want players feeling like they're in a war, not a horrible war - it's this sort of weird war where they can still have fun. So it's an interesting challenge.
GS: Will it be possible to build mods for Team Fortress 2 right after it launches?
RW: In fact we intend to ship the SDK (software development kit) for Team Fortress 2 before we ship the game. People should be making mods before we release Team Fortress 2, I hope. We should be talking about that sometime soon.
GS: Thanks, Robin.