The sheer volume of the Twitch Plays Pokemon phenomenon came as a surprise to almost everyone -- with one little script hacked into a game that was meant to be played alone (and with specificity!), a massive, participatory event was born. Maybe even a new school of gameplay, too. "The developer just did it as an experiment," says Twitch director of business development Brooke Van Dusen. "And then, without realizing it, he kind of happened upon an entirely new dynamic of many people playing one game -- that's an entirely new branch of game design." What if developers began creating games from the ground up to be played on Twitch by big groups of people? What if there were many more games built to be streamed with significant input by others -- changing up the livestream relationship between streamer and viewer, so that the viewer becomes a bigger part of the experience? Van Dusen says it's happening, and that Twitch is looking to encourage developers to work further with its API and SDK to find ideal implementations. It's a natural evolution for the company, which he says has always seen itself as an interactive viewing platform. "People have been trying to make video entertainment more interactive for a long time," Van Dusen says. "We have the largest chat server in the world, and that's what got Twitch to where it is today. You don't only watch video content from an entertainer, you interact with that person. There are a lot of other interesting use cases on Twitch, like people playing Cards Against Humanity, and [letting] the Twitch chat decide whose card was funniest. We've seen this happening in lots of different ways, but it wasn't until Twitch Plays Pokemon that it happened in a software-based way, where someone linked the commands directly."

Twitch Plays Pokemon's 16-day run saw over 1 million people participating -- 121,000 concurrent players at peak -- with over 9 million unique viewers in total. "What it did was prove people are very interested in the idea of many people collaborating on one central game, and playing it in that way," Van Dusen says. And that was for a game that wasn't necessarily originally designed for such an implementation. "We know people enjoy this," he says. "Now there's that justification for doing work that just makes it easier for developers to experiment, and at Twitch we want to support them, if they're willing to spend time creating a game completely for the Twitch platform." At GDC this year, Van Dusen planned to pitch the idea of Twitch Plays to other developers. On the way, he met Michael Molinari, who had the idea already, even before the Pokemon phenomenon had hit. Molinari had been attempting to fund Choice Chamber, an adventure platformer designed for audience participation, through Kickstarter. Upon seeing the game ("this is exactly what I'm talking about," Van Dusen had reflected), Van Dusen decided to ensure Choice Chamber got funded, by encouraging Twitch's popular gaming channels to draw user attention to the Kickstarter and matching donations. It worked. Molinari and colleague Chelsea Howe now find themselves on a new design frontier. One new and interesting challenge involves making a game that scales -- is it fun with a small group of friends on a private channel as well as on a massive Twitch celebrity channel? Smaller groups strategize together, but larger groups build a lot more noise, suggests Howe. "We're looking forward to doing stuff for people to individualize each other in the chat room," Molinari says. "And having some kind of dynamic difficulty adjustment," adds Howe, "if you have, say, a thousand people, you might have a special boss or challenge." Importantly, viewership or number of participants sitting in the IRC chat channel isn't a reliable gauge of how many people are actually playing a game. So Choice Chamber occasionally "polls" its participants to see how many people are actively playing. The game then adjusts for what qualifies as a critical mass of voters for tasks such as summoning helpers. Molinari got the idea for Choice Chamber back in December 2013, after a year of being fascinated with Twitch streams on a daily basis. The advent of the massive Pokemon game stung a bit, he says. "They stole the thunder of having this novel way of interacting by just chatting," he reflects. But ultimately it was validating, and gave his team the confidence to go to Kickstarter in the first place. A playtest with a major streamer resulted a lot of "where can I buy this" responses, which was encouraging, Howe says. Another major design challenge in these types of games is lag. Twitch's delay, says Molinari, is at minimum 20 seconds, or up to a minute or more for some players. "Which means everybody watching the game is seeing different moments at that point, so you can't have precision interaction." Instead, Choice Chamber's polls let viewers make broader decisions, like weapon selections, and remain open for up to 60 seconds, in order to ensure everyone's vote is counted. "That's where all the design for Choice Chamber came from: What things can the player change that don't need to happen just this second? Can they change things around the [streamer], and not influence that player specifically?"

"That's just the tip of the iceberg," Howe adds. "It doesn't take skill from a mass audience to complete a poll. How can you make a small group or a mob have some kind of skill-based challenge... forming a government, creating a strategy?" Choice Chamber's title screen asks everyone in the chat room who wants to play to say hello, and the screen begins to flood with bubbles from all the participants -- an impressive sight when a big stream of players is present. "The immediacy of that response is very satisfying." says Howe. "How can we put more of that throughout the game?" Successful streamers play a key role in how the experience of the game plays out as well -- successful streamers are natural performers, and some have subscriber-only chat, so their communities are more closely curated and collaborative. These large groups of fans can be excited to have their chance to influence and guide the player, often even moreso than the designer of the game does, and it's an interesting challenge for the streamer to navigate not only the challenges of the game, but the psychology of the mob. During a playtest, for example, popular streamer Zeke knew the contents of a Choice Chamber chest, as the audience had voted on what would be inside -- and he didn't want it. He leapt over the chest and left the room, infuriating the voters. "The problem took care of itself," says Molinari, "because the chat, in their fury, only voted on terrible things for the rest of that round... until he promised never to skip a chest again." "The chatters wants to get the most entertaining watch possible out of the streamers, so they'll give that player the most lives so they stay alive as long as possible, but they also want to give them the greatest challenge. Like a torture artist they're turned on by that, and trying to get an epic playthrough every time," he adds. "It's cool to watch the chatters become game designers," Howe adds. "It starts out with everyone being lovey, trying to give the best and easiest options, but it becomes less fun, and they kind of realize that they can make a better game by making the player rise to a challenge. I teach on the side, and part of me likes the game because it's a great educational tool to make people think about game design and game balancing." Talking with Molinari and Howe, it also becomes clear this new type of participatory game presents even a new kind of vocabulary challenge: Who is the "player", the person playing the game on the livestream or the participants making decisions for them to perform within? "We wound up calling them 'streamer' and 'chatter'," says How. "We would be talking about the design like 'oh, players can do', or 'the player should do', and we'd be using the plural to signify streamers." Designing for self-governing societies is one of the most interesting opportunities game designers working on Twitch now have, Howe suggests -- she found Dan Cook's GDC talk on the subject particularly exciting as she sat in the audience knowing Choice Chamber was well underway.

Molinari says nearly any turn-based game would be a good fit for the format, as it removes the timing factor. He hopes to do some experiments and jams on turn-based digital versions of tabletop card games, eventually. He and Howe are considering a Twitch Plays game jam exploring types of play that groups intrinsically have fun doing together. "A narrative game where you're all building a story could be hysterical, or really exciting for smaller groups," Howe suggests. "It's such a ripe space, it's a little bit terrifying." Speaking of terrors, Zombie Studios' recently-released horror game Daylight includes events that can be triggered by Twitch participants. "We actually added the Twitch events in the later stages of development, because we thought it fit so nicely with the game," studio head Jared Gerritzen tells Gamasutra. "The procedural nature of Daylight allows for multiple playthroughs and we wanted to showcase that through media like YouTube and Twitch. Adding the scare events for Twitch was a the perfect fit, and next logical step." "You can play this game completely alone and it will definitely get into your head, [but] Twitch integration allows the viewers to have a say in your experience, which we thought was a brilliant addition to Daylight," he adds. "Adding another layer to the game by letting viewers type in several key words can only add to the overall player experience, and it's thrilling to be able to scare someone by simply typing a word." As Daylight already depends on random elements, it was "easy," says Gerritzen, for Zombie Studios to add scares that audience members could queue up by typing a single word into chat. "It adds to the variety of events that are already happening in the game as you go, but the Twitch commands are also timed, so the actual scares (and game) remain authentic." Twitch's Van Dusen says the company will be working to see what it can do on the tech side to continue to help and support developers making games for the platform. "For example, if Twitch Plays Pokemon had played one ad every 25 minutes while the game was playing, the creator would have made over $100,000 in the Twitch Partner Program," he says. "It's an entirely new business model we could potentially be exploring." "Everyone's been talking about 'in-game ads' for a long time, but this is a little bit of a different application, where a game can trigger an ad with our API and you can build that into an autonomously-running game," Van Dusen adds. Twitch also has a subscription model that lets viewers subscribe to individual channels for $5 a month, allowing streamers to offer different types of experiences to their subscribers versus others. "In general, we think of Twitch as a platform," Van Dusen says. "I don't think this is any different from the way we've been looking at Twitch since day one. Most of our development over the past few years has been creating tools for our developers, expanding our platform reach with console and in-game integrations, and this is a natural iteration on that. It may not have been planned from day one, but it's still in line with what we've been doing from the beginning." "It's still so new that I think the best implementations are still ones I haven't even thought about," he reflects. "At this point, it's, 'okay, what can we try? I know people are interested in this dynamic, so the question is what are different areas we can explore given what we already know about game design?"