The Future of Sandbox Games

Last summer, we launched our new GAMBIT lab, which brings together students and researchers from around the world, to work together to develop projects which stretch our understanding of the medium. Thanks to a grant from the National Research Foundation and the Media Development Authority of Singapore, more than fifty faculty and students from nine different universities and polytechnics in Singapore come to MIT to work with our students, faculty, and staff, in a rapid design and development process. Students working on this project are able to go from conceptualization to user-testing, developing a finished, playable game in a little over eight weeks. Matthew Weise is one of the people we've brought to MIT to help supervise the production process. Weise is equal parts gamer and cinephile, having attended film school before segueing into game studies and then game development. Matt is a producer for GAMBIT and a full-time gamer, which means he not only plays games on a variety of systems but he also completes (most of) them. Matthew did his undergrad at the University of Wisconsin in Milwaukee, where he studied film production before going rogue to design his own degree. He graduated in 2001 with a degree in Digital Arts, which included videogames (this was before Game Studies was a field). He continued his research at MIT's Comparative Media Studies program, where he worked on Revolution with The Education Arcade. After leaving MIT in 2004 Matt worked in mobile game development for a few years, occassionally doing some consultancy work, before returning to work at GAMBIT.

Matt, along with Clara Fernandez and Philip Tan, two other alums of the CMS graduate program who have returned to MIT to oversee the launch of the GAMBIT lab, have been the primary writers of a recently launched blog, affiliated with the lab, which showcases games research at MIT, offers reviews of innovative games, and shares commentary on trends, ranging for public policy to product placements, which are impacting the current games sector. Check it out!

If you'd like to know more about how GAMBIT contributes to undergraduate education in games at MIT, check out this recent story written by Game Tap's Jonathan Miller. Weiss is spearheading the development of a new game, GunPlay, which is featured in the story:

Consider GunPlay, one of GAMBIT's four undergrad research projects this semester. Using the Source engine made famous by Half Life 2, the students are creating a first-person shooter that doesn't feature any weapons or ammunition. Instead, GunPlay is based on the childhood game of guns in which kids used their hands as makeshift pistols and yelled out "bang!" in order to shoot.

While the early prototype is still running in the Half-Life 2 universe, the final coat of graphics will feature children scampering about a playground. The team has already replaced the Half-Life weapons with children's hands. Instead of bullets firing from a pistol, a voice yells out "bang!" You can also switch to a shotgun ("ch, ch, bang!") or a machine gun (bangbangbangbangbang). It's, quite simply, hilarious.

The point of GunPlay is somewhat more profound. First-person shooters are often denounced by politicians and parenting groups for their ultra-violent gameplay. But what if you remove the bullets and the blood and make the exact same shooter with smiley-faced grade schoolers? Is a child shooting another child with his imagination, using the exact same gameplay mechanics as Doom or Halo, still violent?

And then consider what reactions would have been if the team had chosen to take a completely opposite direction that bordered on obscene. GAMBIT faculty member Matt Weise points to Raph Koster, who examined how context affects gameplay. Imagine if, for example, that instead of placing colored blocks in Tetris that you were organizing corpses in a Nazi death camp puzzle game. It's graphic and profane and not a game you would ever want to see. But it is nonetheless an important way to examine violence in games.

It's this kind of dialogue that Weise hopes will get game designers and critics thinking. Is it the act of pulling the trigger that can be considered violent, or does it depend on context, be it the playground, Auschwitz, or the fictional City 17? Answering these questions through research is at the very core of GAMBIT's mission. "Whatever you can do to get people talking and increase discussion on designing games is a good thing," Weise says.

Still, the team behind GunPlay is focused on making the game fun to actually play. Some challenges ahead include creating an argument system for when one Child 1 shoots Child 2 but Child 2 says that Child 1 missed. While this argument was common on the playground when we were kids, how do you incorporate it in a videogame? Use a voting system? A bully meter? Such are the dilemmas for a game designer.

I asked Matt to share with my readers what he's thinking about these days. What follows are his thoughts on Assassin's Creed and Sandbox Gaming.

The Future of Sandbox Gaming

by Matthew Weise

Chris Kohler over at Wired has written a brutal review of Assassin's Creed.

[T]he open-world concept does absolutely nothing for Assassin's Creed's gameplay. I simply can't see any reason why they decided to go this route other than the fact that sandbox games are the hip new thing that all the kids are doing these days. Yes, it's initially very impressive to look upon and roam about this vast, detailed world. But a progressive, linear series of deliberate challenges would have suited the concept so much better.

Sandbox is a term often used but rarely defined. There is a general awareness that the term refers to open-ended game design, but there are many types of open-endedness. In the loosest sense almost any game that does not funnel player navigation into some obvious path could be considered sandbox. The most commonly cited example of this is Grand Theft Auto, with its giant world freely navigatable by car. Recent titles identified as sandbox games often take GTA as a model, as in the case of Spider-Man 2, Mercenaries, or Saint's Row. All these games feature massive worlds, rapid navigation systems for travel, and amusement park-like mission design.

Although Kohler never defines exactly what he means by sandbox it feels like he's using the popular definition, citing the incompatibility of stealth with a GTA-style massive world. "How do you make an open-world Metal Gear Solid? Apparently you don't," he concludes.

Ironically, I've long considered Metal Gear Solid--and many stealth games in general--sandbox games. I've used the term sandbox to refer to any game world--regardless of size and scope--that offers free-roaming, open-ended gameplay. For example, I've always felt Mario64 is the greatest sandbox game ever made because of its ingenious non-linear level design. Levels in Mario64 do not rely on multiple paths but instead allow for improvisational play based on a simple, elegant rule set. In my view this is what all good stealth games do as well. Games like Thief, Tenchu, and Hitman are based largely on open-ended spaces designed for improvisational play. And even once-linear series like Splinter Cell and Sly Cooper seem determined to adopt more open-ended design with each new installment. I'd say stealth has the market cornered on sandbox design in a way no other single genre has... which is why Creed's inability to create a deep sandbox experience is so interesting.

Assassin's Creed's problem is that it's too much GTA and not enough Mario64. The cities are massive and the player can run, jump, and climb effortlessly over every building. This lies in stark contrast to stealth games like Hitman, which also have fairly large environments yet limit the player's actions in ways that create tension and strategy. It's not the complexity of the world in Assassin's Creed that's the problem. The problem is that the player's super-human abilities negate most of that complexity. One can easily imagine a version of Assassin's Creed in which Altair is far more mortal, enemies far more deadly, and climbable structures far more limited. It's also easy to imagine a world where targets are available at all times. Then it would be entirely up to the player to decide how and when to perform a hit--sort of like deciding how and when you get a star in Mario64. In the current game, though, it's as if the designers feared Creed would be too short if players were simply allowed to use the freedom they were given, hence the GTA-style mission design in which the people you are supposed to assassinate only "appear" after you complete task A, B, and C.

When I think of open-ended world design I tend to think of worlds that don't involve such limitations. Call it the result of a childhood playing Ultima. I think of worlds in which, if you need to kill the dragon in the cave and you happen to have a drill, there's no reason you can't just drill straight down, bypassing all his little traps, and kill the bastard. That's open-ended to me. That's sandbox. The pleasure of such incredible agency is much more satisfying than any forced narrative structure.

It's true that Assassin's Creed offers some interesting dynamics. Kohler mentions how absurd it is that your targets--which are really just bosses--will follow you to the ends of the earth in an effort to kill you should you botch a hit. This is sort of stupid, but that doesn't mean it can't result in an interesting experience. Once I was being pursued by a target who had spotted me. He began following me up a ladder to a rooftop, brandishing his sword. Instead of engaging him in a swordfight I waited until his head peaked over the roof's edge, quickly grabbed him by the face, and threw him to his death. I have to admit I found this both hilarious and satisfying. It gets hard after a while, though, to see such character behavior as anything but predictable A.I. since all bosses seem to display this suicidal quirk. And even if there were deeper dynamics to be explored in each boss encounter, you cannot actually go back and explore them without restarting the chapter and sitting through endless unskippable cut-scenes. So while there may be some depth to Creed's dynamics, you'll likely never see them since you've got basically one chance at each boss.

Part of me wonders if a good chunk of Creed's problems might gave been solved by a manual save system. In most stealth games I find myself wanting to perform tasks perfectly, which means I like to replay certain moments over and over in an effort to perform them in exactly the way I find the most dramatically satisfying. Creed actively prevents this: you have to live with your mistakes. While I respect that philosophy in the abstract, I think it undercuts Assassin's Creed. Altair is a badass; the player is not. Therefore the player has to make mistakes in order to assume the role of their avatar. By the time I've replayed a Hitman level six times in order to achieve the Silent Assassin rating, I feel like I've become an assassin. The ability to engage in such trial and error in Creed might have totally altered the experience.

It will be interesting to see how sandbox gaming evolves in the future. Are we going to get bigger worlds with shallower dynamics or smaller worlds with deeper dynamics? Or maybe there doesn't have to be a trade off. I still don't believe sandbox games, stealth or otherwise, have to sacrifice depth for size. Hopefully a game will come along that will prove this someday. Maybe it will be Assassin's Creed II.