The first time games artist James Leggett saw a gladiator fight in VR, it was a simple sword hacking away at “enemies” that were just virtual blocks. But he saw the potential right away — even if it was years ahead of its time. When the team behind it asked if he wanted to work with them, he signed up immediately.
The experience was the very first version of Oculus Maximus, one of the hundreds of small, crude, and often prescient games that were submitted to the very first Oculus Rift game jam in 2013. Developed over the course of a few weeks, there wasn’t much to it: using two motion controllers, players held a sword and a shield, taking on a handful of simply animated opponents armed with fists and, later, their own blades. But the illusion of a massive arena and a cheering crowd gave it surprising weight. (The crudely animated gore, so surreal it became funny, may also have been a selling point.) When the three-person team put up a demo, gameplay videos started appearing online, including one from YouTube superstar PewDiePie. Inspired, they launched a Kickstarter for a larger version, virtual reality’s first fully developed swordfighting — among other kinds of fighting — experience. Then, almost as quickly as the game announced, it was canceled.
Oculus Maximus was drawing on a VR fantasy that people have been imagining since at least 1992, when Neal Stephenson gave virtual katanas to the denizens of his online “Metaverse” in the seminal science fiction novel Snow Crash. Swordfighting — whether it’s with foils, foam swords, faux lightsabers, or wooden yardsticks — is an incredibly intuitive kind of play, and one that would seem to be a natural fit for immersive VR. It’s a martial art, but it can be so stylized that it almost transcends violence, and easily mimicked by almost anyone with a long stick and a nimble hand. And as it turns out, whatever Snow Crash says, it’s incredibly tough to make virtual bodies do well.
Oculus Maximus, which hit the scene early, fizzled under some unique and unfortunate circumstances. Among other things, the Razer Hydra motion controllers it required were discontinued during development — when the team’s small cache started malfunctioning, they were reduced to dismantling and resoldering them. The product that was supposed to replace them never materialized, and VR as a whole moved toward passive experiences. This year, though, the HTC Vive has brought virtual motion back in a big way. It would seem to open up the possibility of acting out heroic fantasies, becoming ninjas, space pirates, and Jedi knights, as well as interacting more naturally with other human beings — reading their body language and responding in kind. But the perfect dueling game — one of the best ways to fuse those two elements — has yet to emerge.
Turn on the Vive, and you’ll certainly find opportunities to hold a sword. ZenBlade, formerly Ninja Trainer VR, lets players slice flying fruit with a motion-controlled katana. The less widely available Star Wars: Trials on Tatooine, a demo from Lucasfilm’s Industrial Light and Magic division, asks them to deflect blaster fire with a lightsaber. As much fun as both these are, though, they’re basically tennis with blades. For actual combat, you can try something like the enjoyable Vive dungeon-crawler Vanishing Realms, but its skeletal enemies’ attacks are still slow, heavily telegraphed, and clunkily animated.
This clunkiness is partly because even outside VR, relatively few gamemakers have figured out how to handle complex and realistic swordfighting. There are exceptions to this, like Kingdom Come: Deliverance, whose intricate fighting system is one of its biggest selling points. But as game developer Kevin Murphy laid out in an essay last year, there are many more satisfying first-person shooting experiences than melee combat ones, and it’s tough to build a realistic-feeling system that doesn’t boil down to either hitting a couple of buttons over and over, or navigating a complicated series of combinations. In theory, VR can help fix that. The headset’s immersive first-person perspective could help players feel more like real-life martial artists, and motion controls could make virtual swordfighting as nuanced and natural-feeling as holding a physical object. As the creators of Oculus Maximus found, the reality is more complicated.
VR dueling starts with a seemingly impossible task: how do you translate a deeply physical experience into one that requires nothing more than waving your arms in the air? Motion-control games can’t restrict characters to a set of polished precanned movements, so gladiators’ virtual bodies in Oculus Maximus sometimes stretched and twisted in uncanny ways. Because the game was reacting to unpredictable sweeps instead of simple button presses, the team couldn’t prepare detailed counter-attacks when enemies got hit — they just flopped back and tried to attack again. And when an enemy blocked a player’s sword, there was nothing to stop the player’s real arm from going straight through the shield.
This last problem — the disconnect between players’ bodies and their avatars — has been an issue for motion-control games since long before VR. In immersive swordfighting, though, it may not be as troublesome as it sounds. Neal Stephenson, a self-professed "swordsmanship geek," offered one explanation while fundraising for an ambitious motion-controlled dueling game called Clang: if you’re fighting well, you’ll be using short, precise movements that make staying in sync with an avatar easy. "Real swordfighters don't make wild, uncontrolled swings without regard for what their opponents are doing," he wrote. His game would use interface cues to nudge players when they failed at this, until they simply developed an instinct for it. Unfortunately, this is hard to test in the crude fighting of most VR games so far.
Clang failed after two years of development, but Oculus Maximus co-designer Eugene Elkin thinks Stephenson was on the right track — and years later, Elkin is still trying to make VR swordfighting a reality. After Oculus Maximus, he was offered a job at VR studio Survios, working on a new action game called Raw Data. It’s primarily a cooperative game that uses a mishmash of weapons to fight waves of robots, but due to a strange design quirk, it also offers a slightly Clang-like experience: when two players enter the game’s lobby, before they start working together, they can fight with fists or glowing blue katanas.
While Elkin describes the resulting brawls as more like "swinging a sword around like a stick," the requisite elements of dueling are all there, and players are apparently willing to maintain the parts of the illusion the game can’t manage. "When you watch kids play when they're doing sword battles, they stop themselves. They pretend like it clashed, because they want it to," says Elkin. As long as players can see their hands and are getting clear feedback like vibration, he says, it’s the same thing in a game. And unlike Oculus Maximus, Vanishing Realms, or any other single-player project, there’s no need to teach an AI how to react to the variability of motion controls.
Even if players are willing to obey interface cues, developers still have trouble designing whole-body VR avatars, since they have to extrapolate an entire human being from two hands and a head. This is only slightly awkward when you’re alone in a game, but if you’re looking at someone else’s avatar, it’s alternately hilarious and horrifying to watch their arms flex like spaghetti or their torso get stuck to a weapon. And it takes away one of the most valuable cues that real martial artists have. "Movement's so important, right?" says Elkin. "You're watching for clues, whether it's anticipation of the body or throwing a punch."
One solution to this, which Elkin suggests, is to add more motion trackers, something that could be done with full-body suits or depth sensors like the Kinect. Another is to use machine learning to better predict movement based on the precise angles and positions of someone’s headset and controllers. And yet another is to find a way to simplify the avatars, so that while players might not feel exactly like they’re fighting a real human, they’ll get a stylized target that moves and thinks like one.
For Raw Data, the biggest problem is a mundane one: lag. Right now, player-versus-player battles are only feasible when the players’ VR rigs are in the same location, connected over a local network. Online, even a little bit of delay will throw the action out of sync, so one person is raising a sword to block when the other has already landed a blow. This could potentially be fixed by making everything in the game move slightly slower than in real life, something Stephenson also planned to do in Clang. But multiplayer battles aren’t a focus for Survios, which is still building the cooperative section.
Many people who try VR won’t have any interest in dueling with friends or an AI opponent. But just like action-packed first-person shooters helped pioneer the mechanics that would show up in smaller narrative games like Gone Home, cracking swordfighting could open up all kinds of new VR experiences. It involves dealing with the same problems that underlie every attempt to let people physically engage with other creatures in VR, instead of simply talking to them, running around them, or shooting them. That includes other martial arts like boxing, but also experiences like dancing, hugging, or even high-fiving. To get truly close to somebody, you might need to learn to fight them first — and in VR, we haven’t figured out either one.