Skip to main content

Speedgate, the sport invented by artificial intelligence

A design team’s collaboration with an advanced neural network produces a surprise success story.

By Glenn McDonald

Shall we play a game?

Computers and games have enjoyed a wildly successful partnership for decades, and now comes word of a new collaboration: A field sport that combines elements of soccer, basketball, and Ultimate Frisbee. A game with built-in ethical rules from a curious source. A game invented by a computer.

Earlier this year, a team of designers in Portland, Oregon, tasked their artificial intelligence system with building a new team sport for us humans. The result of that pet project, Speedgate, is being heralded as the first viable AI-created field game. Speedgate has since become an indie sports phenomenon, with proposed leagues in more than 50 countries worldwide.

Speedgate got big fast, but its creators didn’t start with serious ambitions. It began when a small group of designers at AKQA, a creative agency office in Portland, were trying to come up with a project for a local design festival.

“We all love sports; one of our big clients is Nike,” says AKQA creative director Whitney Jenkins. “We thought: What if we built an AI model and tried to create a new sport?”

The designers wanted a game that was fun, easy to learn, and accessible for a broad range of players. So they turned to a commercially available AI system called a recurrent neural network, or RNN. This is the same kind of AI that programs robot control systems and speech recognition programs, AI that quite literally teaches itself through repetition and trial-and-error decisions. The programmers trained the system by feeding it information on more than 400 games, from basketball to biathlon, cricket to cornhole.

The AI mixed and matched elements and made suggestions, spitting out concepts and phrases in basic descriptive terms. Team members then translated the suggestions into natural language and fed the ideas back into the machine. With each round of translation and processing, more details took shape.

“For instance, the AI might say: ‘Game where for to hand ball throw two players jump and kick,’” Jenkins says. “You take the output and tell the computer: ‘Say it like this: A game where two players kick and push a ball.’ You give that back, and the computer starts thinking that way. Eventually, the AI starts making suggestions like: ‘A game where anyone can pass and kick.’

Eventually the AI started to generate “sport-like ideas,” Jenkins says with a chuckle. “One of my favorites could be described as a version of tennis: Two players hitting a ball back and forth, but they are balancing on a tightrope that’s being elevated into the air by two hot air balloons.”

The system also suggested an exploding Frisbee game and a kind of underwater parkour. After weeks of data-crunching and translation, the designers narrowed the field to 10 games, test-played them at a local park, then narrowed them down to three. Finally they arrived at the best version of the best idea: Speedgate.

Played by two teams of six, Speedgate is a fast-moving game in which players throw or kick a rugby training ball to teammates, but can’t run with or carry the ball. The objective is to move the ball through a gate in the center of the playing field, which then opens up that side of the field for a scoring opportunity. The team then aims to kick or throw the ball through another gate at the end of the playing field, which can be of variable size, depending on space.

Here come the twists: The gates at each end work like soccer goals, except that there is no net. So teams can score from either side of the gate. A team wins bonus points if it scores a goal, then immediately passes the ball back through the gate — a play called a ricochet.

Somehow, the AI seemed interested in fair play.

AKQA has set up an interactive online tutorial that highlights other unique qualities of the game, each generated by the AI. For instance, if a team claims to have scored a goal or rebound, but actually didn’t, then the points are awarded to the opposing team. It’s one of several elements the AI introduced that appears to promote ethical play.

“We haven’t really been able to dig out where it got these sportsmanship elements,” Jenkins said. “But somehow, it seemed interested in fair play.”

Speedgate is a fast-moving game in which players throw or kick a rugby training ball to teammates, but can’t run with or carry the ball. Photo from Speedgate

Andrew Wilson, a 22-year-old electrical engineer in Phoenix, read about Speedgate earlier this year and started a weekly game at a local park with friends from college. He said the game’s seemingly random assortment of rules and gameplay features actually clicks together nicely.

“The very first time that I actually played the sport, I was honestly surprised at how fun it was,” says Wilson, who was initially just curious about the AI aspect. “I thought that it would have all kinds of weird rules, but, surprisingly, it was really well-thought-out. It’s all about teamwork. You have to play the game as a team. Otherwise, you’re going to lose.”

That need for teamwork may be the most popular element of Speedgate, says Jenkins, who has received calls and emails from teachers who want to teach it to their elementary school students.

“Because of the central gate and the omnidirectional movement, it helps kids think about problem-solving in new ways,” Jenkins says. “And it’s not a contact sport, so kids aren’t going to get hurt.”

Since Speedgate’s debut in April, people have started active leagues in several countries, including Canada, Australia, France, Iceland and the U.K. In addition, AKQA has received requests for official rules from more than 50 countries total. Jenkins says his design agency is now in the unexpected position of promoting an alarmingly popular new game that’s generating its own momentum. Is AKQA going to get into the professional sports business?

“We’re working on it,” he says. “We did not intend to create a new global sport, but we’re thrilled that it’s worked out the way it has.”

Jenkins says it’s important to note that, while the game was technically “invented” by artificial intelligence, it’s really more of a human-machine collaboration.

“AI never could have, on its own, come up with a fun, practical, playable version of Speedgate,” he says. “But on the other hand, I don’t think we could have ever come up with this without the AI.”

Jenkins says his experience with Speedgate, and its remarkable viral success, has convinced him to keep working with non-carbon-based talent on future projects.

“In a way, AI became a member of our creative team,” Jenkins says. “It was a valuable thing because it thinks fearlessly. It presents a million ideas — some great, some ridiculous — but there’s no fear attached; there’s no self-editing. It provided us with options we just would never have thought of.”

Published on

Glenn McDonald is a writer based in Chapel Hill, North Carolina.


Illustration by Gary Neill


On this road trip, the car wrote a novel

Retracing Kerouac’s route, a gonzo data scientist created a surreal work of AI-driven fiction

By Tony Rehagen