To get an edge on game day, many athletes will consult opponent statistics and game footage in an attempt to recognize patterns. This human-powered, predictive analysis forms the basis of a team’s overall strategy—but the human brain can only spot so much.

To help players understand more precisely how they should act (or react to opponents) on the field, pitch or court, sports organizations are turning to artificial intelligence and machine learning tools.

Just like that, teams can add a near-perfect eye to their game-day preparations. The growing number of tools that leverage big data, artificial intelligence and a robust network offer a game-changing opportunity for sports teams looking to improve performance.

Big Data Takes the Court

Big data isn’t anything new in the sports world, but as data gets bigger, teams—like any company—continue to play catch up in order to decipher it.

Early examples of this data-mining AI used to analyze performance have been prevalent in professional basketball for years. In 2013, 15 U.S. teams purchased a camera-tracking system to record every movement players made on the court. Those movements go back to the coaching staff as geometric coordinates, and with some coding work, analytics teams can transform the coordinates into playable video files that coaches and players can learn from.

The problem with this and similar technologies is the time all the data processing takes. To mitigate the issue, the company behind the camera-tracking system is in the process of giving their system an upgrade. Using deep imitation learning and player tracking data from a professional soccer league, researchers recently developed a unique “ghosting” system that could change the way athletes and coaches approach training. Deep imitation learning enables a computer to learn multiple things from data points without a huge expenditure of human engineering effort.

In the sports world, this will allow coaches to measure their own players’ performances using “ghost” players that first track and monitor what a player did on game day, and then, in a digital simulation, behave according to what that player should have done in the same scenario.

Deep imitation learning “avoids the need for many-years of manual annotation,” wrote the researchers who created the camera-tracking technology. “Our ghosting model can be trained in several hours, after which it can ghost every play in real-time.”

The Next General Manager: A ‘Big, Artificial Brain’

Deep imitation learning creates a situation where data points analyzed using AI can be dumped into a system that not only creates a variety of play scenarios, but also learns on its own, over time, how to adjust and manipulate that data according to new variables.

Making a system like this work is easier said than done, though. Teams interested in deploying a similar ghosting system came up with will have to make a serious investment in robust computer networks, with a cluster of computers and processors—something that Shelly Blake-Plock, CEO of Yet Analytics, says works like a “big, artificial brain.”

“You have to have something that is robust enough to constantly be able to read through data sources in accordance with what it’s seeing,” he adds.

Teams that can implement this technology effectively will have the advantage in game-day preparation, fielding players who have the brainpower to match their desire to win.