GI: Why did you change up the fundamental Team Fortress gameplay of capture the flag with a majority of the maps?

Walker: One of the main reasons why we've gone to control points with much of the maps is control points are about defending and attacking points, and TF's gameplay, all of the classes fit into some part of a defensive or offensive role. It solved a core problem with Well and CTF where if the flags are at each end of the map the only bit that's really important are those areas. So if you're a defender, Engineers don't build sentry guns in the middle of Well or three quarters into Well, they build them at the end because people will walk around your sentry gun and steal the flag. So we were looking at half of this map that was hardly being used and if you're an offensive class you want to avoid combat in the middle of the map to be at full strength when you reach the enemy base. Our game should be about getting into combat and having fun, so we don't want to create a situation where people are trying not to get into combat, or they're trying to avoid combat.

So, with the capture points, if you're an Engineer, in the flag maps you build sentry guns at the flags, you never move them, that's where they are. You might move them if they get blown up. With control points there's only the front control point that's in danger, enemies can't take control points behind you, so Engineers end up, you'll build around the control points, and as you capture control points you move your sentry guns up. Take another one, move the sentry guns up. Move your teleporters and so on. For the defensive classes going to control points meant they saw a lot more of the map. Defending the middle for example as an engineer is significantly different than defending the end, so Well as a defensive class has a lot more variety in its play with control points instead of flags. So that's some of the theory in why we went with that.

Brown: I still think having the control point scheme, and having the flags at the bottom is a real easy way to tell the player where the front line is too. So as far as when you spawn and when you get into combat and where you can expect to interact with other people will be pretty obvious from the get-go.

Walker: Yes, if you join a CTF map, if both your flags are home you have no real understanding what the state of the game is. Maybe you should be defending because all of the enemy is outside your base, they've pushed your team all the way back, or maybe it's vice versa. Whereas if you jump into Well, you can see, well we own four of the capture points and they own one, we know we've pushed all the way back to their final capture point already. It's a great way of exposing game state to you in this very clear way. It tells you as a player who's jumping in, here's where to go. Oh, I'm an engineer, I'm going to defend the furthest point we own. I'm on offense - I want to attack the first enemy point.

GI: So, we could still see something like Murderball, Push, Team Sniper, etc.

Walker: Yeah, I think with all those we're going to play test and evaluate them all. Our focus, perhaps this is a reflection of our learning process, early on with TF2 and TFC I don't think we knew what was good and what wasn't so good. We had a shotgun approach. Let's just do a bunch of stuff. Some of those worked out and some of them didn't. So our goal here is to lead with our best foot first, and start experimenting a little later.

GI: What about people who are going multicore? Are you supporting that as well?

Brown: Yeah. We've already moved a couple of systems over, and definitely the performance in general increases, but what we can do with the particle systems and stuff like that will also increase. So the systems that we've already got in place are the particle systems that are multi-cored, and you saw what we can do with that on those machines is much greater. Animation-wise, not that we're doing a lot different at the moment, that system has been offloaded, so just as a general performance thing, it performs.

Walker: That's actually the most expensive thing we're doing in our character set. These characters are more expensive bone wise and animation wise than we've ever done before. Just getting the characters, the characters have so much stylization and personality in their models that we wanted to come through in their character animations as well. It took a lot of work, and as a result, performance wise they're more expensive than we've ever done before, character wise. It's definitely more expensive than DoD [ed: Day of Defeat] and so on.

Brown: It comes out a lot in the high end.

Walker: Yeah, if you've got a multicore machine, it helps a lot. As I say, the quadcore systems that we have here, their frame rates are just insane. It's like 300 FPS at 2500 by 1900 or something like that. Those cores are just chewing up the animation systems and particles. All of our core technology systems, the engine itself is shared between TF2, Portal and Episode 2 and so on, so all the work that's being done there is being shared across all of those products.