By John Walker on November 8th, 2007 at 4:27 pm.
Coming up in December’s PC Gamer is my interview with Half-Life 2 Episode 2’s project lead, David Speyrer. And as is often the case with interviews, there was a lot more said than could fit in the magazine. As a teaser for next month’s Gamer content, here’s some bonus discussion about what it’s like working for Valve, how play-testing impacts on the games development, DX9 vs DX10, and the role of consoles. Spoilers a-go-go.
RPS: Could you talk about your role in the Half-Life universe?
David Speyrer: I interviewed on the day that HL1 went Gold, and started early in the year in ’99, at the very beginning of HL2. I was on the project for the whole duration. Towards the middle of the project, we formed into four cabals – mini design teams – for design and production, and each team was responsible for building a section of the game. I ended up as the cabal leader of the team which made Canals and the Citadel. I was a programmer. I worked on the air-boat, the poisoned zombie and a bunch of other things. Then on Episode Two I’ve been on it from the start as the project lead and programmer. I did specific work on the Hunter AI and some of the car stuff.
RPS: What do you see are Valve’s motivations to continue making games? Surely money can’t be a big part of it any more?
DS: That’s not really true. We have a lot of employees. We’re close to 150 people, I guess. We have to pay everyone’s salaries. Also, this is pretty much what everyone wants to be doing anyway. We’re kind of stuck. We’re just going to keep on making games. I think everyone has a strong interest in the next thing. And when the next thing becomes the current thing, everyone develops an even stronger interest to see it through to its potential, and get it finished and out the door. And then there’s always a period of collapse at the end where half the team is saying, “I can’t even think about making games right now”. And then a month later, everyone gets really excited about what to do next and what gameplay experiments they wanted to run and the thing they always wanted to see in our games… which builds a momentum of its own. It’s what we all love to do. It’s such a special company to work at. Everybody feels pretty lucky to get to do it with these folks.
RPS: I recently replayed Episode One on the Hard difficulty setting, and found that while more challenging, it didn’t have quite the same flow. Playing Episode Two on Normal, it once again had that steady, continuous sense of progression and challenge. How do you go about trying to deliver this?
DS: We play-test every week with an external player throughout the development process. Sometimes we’re doing it maybe twice a week. The whole team watches every play-test. We’re really looking for places where people are dying a lot and reloading a lot. What you’re really seeing is just the result of hundreds of hours of observed play-tests. The high level goal which we keep in mind is that dying a bunch in the same place isn’t very fun. We absolutely saw in Episode One, through our Steam stats gathering set, there’s a strong correlation between a spike in player deaths and players quitting and not picking the game back up. That made us really aware of the danger of a difficulty spike in your game. We invest a lot in our story and the game as a whole, and we want players to get all the way through the game, we want them to get the most out of the game and to see the end an be satisfied by it. We work hard when we see a problem in the play-test to smooth those out.
RPS: Have you any examples from Episode Two?
DS: The solutions to the problems can be as disparate as the maps themselves. There were the tunnels where the Ant Lion guard – the cavern guard – is chasing you, and you’re trying to get to the larval extract. That was one of the early places where people died a lot. In that case, the solution was to add a bunch of clear landmarks so players could tell where they were. It really is a bit of a maze through there. We added a stronger visual design, more clear landmarks, and we simplified the maze so there was a more clear forward direction, so players could know they were getting nearer the end. Initially it was very circular with a lot of loops, and players would often be going back towards the beginning with no idea. We saw players get stuck in that level for over an hour, which was very frustrating as the whole time they’re being chased by a monster which is killing them over and over again. Another place people died a lot was the last map. That was largely owing to all the options we present for how to fight that battle. There were ways of attacking that final fight we hadn’t considered in the design, but became obvious when we watched play-tests. We would have to watch sometimes four play-tests or so on a map before we’d see patterns start – like there’s a whole class of player who doesn’t know about any of the alt-fire.
RPS: The big hint was handy for me there. I hadn’t realised that it was the ultimate way to destroy the Hunters.
DS: I’m glad you caught that. A lot of players miss that. So we were initially relying heavily on the alt-fire mechanic for beating that map, and we steadily had to introduce other means for different players. There’s a lot of logs around you can pick up and throw at Hunters, for players who know about doing that. There’s also rockets around you can use to pick off the Hunters for players who know about the RPG. Then there’s the [Overwatch Rifle’s] alt fire ammo in that map. And you can drive around in the car and run over the Hunters. That’s another way.
RPS: Those damned Hunters. It was a strange moment – you’re finally given this incredible way of wiping out the uber-bads, the Striders, and then the Hunters do their best to stop you using it.
DS: It’s this whole tug of war between you and the Hunters who are protecting the Striders. But tuning that map was mainly about providing a whole palette of options for players with different play styles. That was kind of a unique problem for us in the franchise. We’d never given players so much freedom to fail before.
RPS: If I had a cricitism of Episode One, it would be the formula showing through a couple of times. A number of sections were three-stage puzzles, or three-part tasks. But this didn’t seem to occur in Episode Two.
DS: We try to make our training feel as natural as we can. In the Citadel of Episode One, you’re dealing with such abstract mechanics at times – the Combine balls, the grabbers and the light bridges and all that stuff – that it’s kind of hard to express training for those scenarios in any format other than doing them. Episode Two’s setting being very natural and pretty real-world – there’s not a lot of wacky sci-fi stuff in Episode Two – so we were able to hide the training. And we spent a bit more time building Episode Two as well, so what you’re seeing is we were able to smooth out the rough edges of our training and move some of it into entertaining character dialogue and things like that. Maybe it’s just a function of spending more time.
RPS: You chose to not use DX10 for the new games. Why was that?
DS: For this set of products we decided to use some DX10 features, but not through the API itself. It’s because we didn’t want to be Vista only. We didn’t feel there was an enough of an install base for the Orange Box launch. I’m sure we will use it – it’s pretty much inevitable. But this time, things like the hardware face morphing is implemented via a back-door API.
RPS: DX9 can do what DX10 offers then?
DS: Yeah, but through driver-back doors. And I’m sure we’ll rethink it at some point. For this roll out it was fine to do it as we did.
RPS: The Crysis team said they can’t do their day/night cycles in DX9 and they need DX10. [This question was asked before the DX9 hack for the Crysis demo was uncovered] Would you say it’s not true?
DS: Not for the features we’re doing now, but maybe down the road. Maybe even for Episode Three, there may be something which makes it mandatory to go after. We pursue that stuff on a case by case basis, purely on whether it gives value to our customer. What do our customers get when we make this investment?
RPS: Has including consoles in the Orange Box development changed how you go about things?
DS: No, not really. Other than how it affected the schedule, which was large. The simultaneous launch on the PC and X360, with the PS3 lagging a little behind, meant we had to be finished well in advance of when we’d normally be finished for a PC launch. The time between ‘gold master’ and ‘street’ is pretty short on the PC. But with the whole certification process on the console, that can’t be the case. The biggest impact was it creating a really odd shipping cycle for the team. Normally we’re accustomed to working really, really hard, going gold, and then having a short period of time before everyone is playing the game, with everyone really exhausted from the shipping push. It’s really exciting. This time we had a bunch of people on the team who had’t had anything to do on Episode Two, or the Orange Box in general, for quite some time, ever since the 360 build went into ‘cert’. Because when the 360 build is done, you’re not going to substantially change the game. It meant you were happy with the console product, and our goal was always to have the 360 product to be first class, of parity with the PC product. If we’re happy with that, then by definition we’re happy with the PC product. It was a pretty relaxed shipping cycle after we hit cert, but going to certification was a huge push for the whole company. We had people play-testing in shifts around the clock getting to certification.
RPS: The other thing that stood out was quite how funny this new episode was. The introduction of Magnusson, and oh God, the microwave joke. Looking at the whole of the Orange Box, it’s kind of a theme. Was this intentional?
I guess that I hadn’t really considered the fact that all three games in the Orange Box have a pretty good dose of humour to them. I know in Episode Two we did really want to bring more humour into the franchise. I mean, Half-Life had plenty of humourous moments, if a dark kind of comedy. The setting and premise of Half-Life 2 was pretty dark, and we did try to make you believe in the Combine oppression of humanity. But over time, as a team, we wanted to bring back some of the Half-Life humour into the franchise without diluting the sense of gravity. There’s a lot on the line for humanity and there’s real consequences for major characters in the game. Once again, back to the contrast thing, having the humourous moments makes the dark serious moments feel more impactful, I think.
RPS: I think you’re right. It gives you a greater scope for believing in the character’s emotions. When Eli told me he was proud of me… I felt a bit daft. I felt so moved!
DS: I react in the same way to that line. It means a lot to me when I hear him say that. Also, Chet and Eric Wolpaw are writing here now. The Old Man Murray guys. They inject a lot of great humour. Eric did pretty much all the writing for Portal, and that dark humour really comes through there.
RPS: It’s a shame that Episode Two can’t end with a song.
DS: Yeah, that song is great. I also really liked, “You’re the fastest subject to ever kill your companion cube”. And then you feel bad. Was I really the fastest? C’mon!
RPS: A trailer for Episode Three was notable by its absence.
DS: We deliberately left that out for a number of reasons. One was we didn’t want to dilute the moment of the ending. I think if you’d watched the credits roll and then this high action Episode Three trailer came on, you wouldn’t quite feel the same. Another reason was to leave us open on Episode Three. We’re going to try and do something pretty ambitious for that project. We don’t want to over commit. If you look at the Episode Two trailer that we shipped with Episode One there’s some pretty radical difference between what you see there and see in finished game. That’s really an artefact of making a trailer for a product that’s still in heavy production. You just don’t know where you’re going to end up.