Best Of GDC: The Secrets Of Portal's Huge Success

[Continuing Gamasutra's 'best of GDC' show follow-up, we go in-depth with Valve's Kim Swift and Erik Wolpaw at the Portal postmorterm, examining the “iterative process” behind integrating story and gameplay in the 2007 Game Developers Choice Game Of The Year.] Despite holding a session in a room full of clearly die-hard Portal fans, Wolpaw and Swift opened by warning for “spoilers” and expected the audience to query: “Why should we care about Portal?” Introducing why, Swift said, “We had a really small team, about 10 people,” and for those who wondered why they should care about how the team chose to integrate narrative and design, Swift opined that “By itself the story wouldn’t make much of a novel (don’t get me wrong, the dialogue is pretty fun), and on its own the gameplay would be fun, but kind of dry. But the tight integration of out story and gameplay resonated with people.” Wolpaw introduced the “crackpot theory” which led to this integration: The feeling that games tell two stories: the “story-story” and the “gameplay-story.” “If you lower the delta (the closer you bring them together)," he said, "it makes the game more satisfying.” As an example of a game with a high story delta (a game in which the “story-story” and the “gameplay-story” don’t intergrate) Wolpaw gave the example of Clive Barker’s Undying. “As I’m in a room full of developers it would be a super dick move if I run down a lot of you, but this is a good shooter and regardless of what you think about his style (that’s one super scary pair of pants) he’s a good writer,” Wolpaw quipped. He therefore gave the example of a game where the player is constantly under threat from attack by enemies, but features many cut-scenes where the protagonist calmly chats with other characters without ever “grabbing them and screaming ‘get out of the house! There are monsters everywhere!'” So Portal’s narrative design goals were to have a “story-story” that never intruded on, or contradicted, the “gameplay story.” And to do that, less is more: “we were ruthless about trimming the fat.” “Playtesting is probably the most important thing we did on Portal,” Swift continued. “Sit down and watch people play your game. Don’t just have them send you reports: It’s not going to be as valuable. You can find out what you’re players actually want by watching them. Adjust gameplay to what players look like they need, and adjust story to enhance what players are already feeling.” “At the end of a playtest we’d ask players to tell us the story back,” said Wolpaw. “If they couldn’t tell us, that was a real sign that they weren’t paying attention. “ Playtest Early and Often Swift revealed they began testing the first room from the moment they started at Valve. For example, the first room initially featured a “shimmery force field” and players didn’t understand what it was -- so they changed it to glass. Throughout the game they proceeded to make the world cleaner and simpler to help teach the players what they need to know. The “color” came from the dialogue, and Wolpaw had some advice for those in the room writing a funny game: “God help you. It’s unpleasant.” “Tough guy dialogue is just about as macho the 50th time you hear it,” Wolpaw explained. “Funny dialogue is funny once -- maybe. A few years ago I worked on a game called Psychonauts. At Double Fine everyone sits in one big room and for the two and a half years I was there I had to sit there in the middle of this pit hearing my supposedly funny dialogue blaring out of monitors and being greeted with stony silence. It was torture -- for me. So if you’re writing a funny game, be prepared to sit in a dark cold despair during development.” He did, however, ask the audience to not despair, as long as they trust their instincts and “remember the initial reactions” the pain could be alleviated. “Evolve the narrative out of gameplay,” Wolpaw advised. “Write to enhance what playtesters are feeling -- keep the story ‘wet’ and don’t get too attached to anything.” “Gameplay too,” Swift continued, recounting the story of the “box marathon” level. Initially a level where it was easy to destroy the box, they changed it to a level where it was impossible to destroy, and featured more gameplay events which featured the box (to help the player to remember to keep it.) They tried to remind players by making sure they could always see the button at the end of the level which would require the box, and tried to put hints into the environment, but when all else failed... “I’d been reading these declassified government interrogation manuals, and one of the things that they talked about was that isolation leads people to become attached to inanimate objects, and maybe is GLaDOS needled you a bit you would become attached to the box,” said Wolpaw, explicitly describing the creation of the “weighted companion cube.” Gamers might be surprised, but Wolpaw revealed that the end of the level was, initially, that the player just had to leave it behind. However, “forcing the player to incinerate the box was a great satisfying ending.” As at the time they were also struggling with developing the final boss battle and discovered that not only was this satisfying to the story, it was also an excellent way to train the player for the final boss battle, allowing the mechanic to later work as revenge. The Boss Of It All The boss monster for a puzzle game: Isn’t it obviously a complex puzzle? “The first puzzle that we tried was James Bond lasers,” Swift revealed. “The lasers were extremely boring to dodge and really difficult to aim, and it was really hard to tell if you were hit. So we abandoned this idea and decided to use rockets.” The second idea that they had was “Portal Kombat,” a high intensity rocket battle with GLaDOS in a room full of turrets. “The high intensity gameplay sucked,” jumped in Wolpaw, “it sucked so much that I’m not going to let Kim talk. No one paid attention to what GLaDOS was saying. There were some hard core shooters who got some fun out of it but it alienated the people who were enjoying the puzzles.” The final attempt that they tried was a chase sequence, chasing GLaDOS along a corridor. “The pacing was just horrible,” Wolpaw lamented, “Players didn’t know where they were going and they’d wander around lost … and once they actually got to the corridor we wanted to spring a trick on you, so we had all these pistons spring out of the walls, and it failed in every way that’s possible. Bad pacing and players didn’t really know what was going on.” So, as Swift explained, by this point all they had learned was a complex boss battle was only going to slow the player down, confuse them, and lead to bad pacing. “It’s funny now,” Wolpaw deadpanned, “but we were screwed, because Episode 2 was winding up. In the end, our old friend playtesting helped us. There was one piece of feedback that helped us that there was one part of the game that the players found very satisfying, and that was the fire pit. They were consistently telling us that this was not only dramatic and exciting and it was a really tough puzzle.” “That made absolutely no sense," he added, "When it comes right down it it, that’s about the easiest puzzle in the entire game.” Swift: “We had to work out what made this puzzle climactic? There was time pressure, and a high visual impact.” Wolpaw continued, “It was a dramatic high point, because for the first time GLaDOS is pretty obviously trying to kill you, and it’s first time you can actually control your environment to really escape her.” So for the final battle with GLaDOS, they “just added a timer counting down until a neutotoxin killed you.” “Putting a timer meant I only had to write 6 minutes of dialogue, compared to all of our previous final puzzles, which had required me to write an infinity of dialogue,” Wolpaw joked. Describing the development process that led to Jonathan Coulton's internet song sensation Still Alive, Swift said “We wanted players to leave the game genuinely happy and with a smile on their face,” with Erik expanding that they “we wrote down a list of what would make people happy, and a song kind of floated to the top.” “When all is said and done a lot of this came down to our constraints,” Swift admitted. “We couldn’t create a massive FMV ending of doom.” Wolpaw countered that “Without constraints Portal would not have been as good as it tunred out to be.” “Have faith in your writing and in your team (they’re really the best asset you have),” Swift concluded, “and playtest, playtest, playtest.”