28 Jan 2016

Last week, we touched on virtual reality in sports and howfor players and fansVR could totally change the game. But when it comes to infusing technology into one of America’s primary pastimes, it’s not just about altering the way spectators watch the game and athletes relive it. Using artificial intelligence, we could completely change the way the sport is called from the sidelines as well.

We’re talking about making the computer the coach.

Before we get into how AI could be used to potentially (and, it looks like, successfully) coach a game, let’s talk about the tech in general. Actually, let’s talk about the fact that everyone is talking about AI, and the fact that it’s a type of technology that’s being invested in more and more heavily and focused on in a variety of manners by all kinds of tech heavyweights.

Google, for example, in an effort to make their artificial intelligence more…intelligent, open-sourced their platform, TensorFlow, last November. The company claimed that they “hope this will let the machine learning community—everyone from academic researchers, to engineers, to hobbyists—exchange ideas much more quickly, through working code rather than just research papers. And that, in turn, will accelerate research on machine learning, in the end making technology work better for everyone.” Meanwhile, in December we discovered that Elon Musk and other investors dropped a cool $1 billion to create non-profit OpenAI, which is focused on the development of, what else, artificial intelligence. Musk and co. also promised to open source all of OpenAI’s work to facilitate the growth of the technology (and, perhaps, prevent what Wired dubbed an AI apocalypse). Then, just this week, Microsoft open sourced their AI framework. And it’s not simply a type of technology relevant to developers and data scientists; odds are you interact with the technology every day, whether you know it or not.

So how does this relate to the world of sports–and specifically football? Well, it’s far more advanced than Nick Saban snarling at Siri or Bill Belechick asking Alexa where he can find some (allegedly) deflated footballs. Last week, Wired and Sports Illustrated released a video on how artificial intelligence could potentially be applied to play-calling, allowing coaches to overcome their emotions and instincts and instead rely on real data to determine which plays to run.

The video touches on the (extremely successful) strategy of Kevin Kelley, a high-school football coach in Little Rock, Arkansas, who wanted to base decisions on data rather than his gut. He looked at real statistics and started making play calls and changing the way he coached the game to reflect that info. Instead of operating on his feelings, he looked predominantly and objectively at his options, which were determined by data. Fast forward to the present, and a team that had previously never been to the final four in 40 years now has five state championships.

Now that being said, the idea of using data to make coaching decisions is nothing revolutionary. Every time a coach makes a decision based on basic information like his kicker’s capabilities or the weather, that’s data. Meanwhile, algorithm and machine learning isn’t brand-new to the world of fantasy football. Take Wired-featured Swish Analytics, which describes itself as a “machine learning system for sports betting and fantasy,” and claims their “data science team leverages the most advanced data processing technologies to consistently iterate on millions of sports-specific algorithms.”

But what about giving the coach access to that same technology, and taking away the instincts and emotions and other human factors that come into play? What if the coach had access to thousands of data points instantaneously and an engine that, based on that real-time info, could generate the best play call so the coach is able to make decisions without emotions and instead rely on data. We’re talking about a super computer as a coach.

But what does that really mean for football?

On the one hand, Kelley brings up the point that better, computer-run play-calling could lead to more completions, more touchdowns and a higher scoring game. (And there’s certainly no denying the fact that fans love to see points on the board.) But what about the authenticity, the instincts, the risk-taking? The extreme highs, heartbreaks, and the true experience of the players, coaches and fans? Are we taking a naturally emotional act and making it borderline robotic?

Maybe we don’t have to choose between the computers and the human element.

“I think integrating AI into coaching is inevitable to some extent,” said Chaotic Moon Creative Technologist Philippe Moore, “but I don’t think that removes the element of individual decision-making on the field. Peyton Manning, for example, is so successful calling audibles and making decisions at the line of scrimmage based on his experience, what he knows about the defense, etc. He knows what play or what action is going to give him a better possibility of succeeding. Even if a player has a computer telling them what might statistically be the best play for success, there’s still a possibility that could change in the moment.”

In other words, just because a computer says that, based on data, X or Y play is the best to run in this situation, it can’t take into account every variable. And human performance is a very big variable.

Another interesting angle to consider is, instead of basing decisions on the AI, betting against AI. For instance, an offensive coordinator might think, “Hey, AI is clearly saying this is the best play, and the defensive coordinator is going to expect it, so instead we’re going to take a risk and do the unexpected.” It becomes almost a betting game: Should you do what’s more likely to work? Or will the other team expect you to do what’s more likely to work? Rather than make things predictable, in some ways, it could actually encourage coaches to take more risks.

And it’s not just a computer as coach that we’re talking about. Take, for instance, referees. That’s a position in which your job is actually supposed to be black and white, and it’s hard to see a downside to finding a way that ensures a game is called more fairly. This season in particular was controversial from an officiating standpoint, and if a computer was capable of making more accurate calls…well, a lot of mistakes could be avoided.

In the end, it’s not whether or not AI is integrated into football; it’s just a matter of when, how and to what degree. When we look infusing the sport with artificial intelligence, there’s little to no doubt that the technology will come into…um, play…more and more in the future.