Expert reveals how AI is set to transform the AFL and change how game is coached

Artificial intelligence is creeping into the AFL, begging the question could computer coach an AFL team one day? SHANNON GILL explores footy’s biggest change.

Champion Data founder Ted Hopkins once told colleagues that one day they would “build a computer that could coach a game.”

For now, the company’s Executive Director Tim Kelsey says Chris Fagan and John Longmire need not worry about their jobs being swallowed by Artificial Intelligence bots.

Yet change is coming.

“At some point a computer using artificial intelligence, would be able to think about what’s going on in the game and what it might do to counteract what’s happening,” Kelsey tells CODE Sports.

“But people get nervous when you go down this track – ‘Oh it doesn’t allow for your own thinking’ – so we talk about data-informed decisions for a coach, rather than data-driven decisions.”

“A coach needs to be informed and needs to also use their own Judgement in today’s game.”

Dean Cox, Game Strategy & Ruck Coach of the Swans instructs players on a laptop in a pint to the future of the game. Picture: Michael Willson/AFL Photos via Getty Images

In recent months a Matrix-version of AFL football crept a little closer when Champion Data, who provide data to all 18 clubs and the league, got serious about the use of AI in its operations.

Kelsey will be speaking at the Australian Sports Tech Conference at the MCG this week, and sharing some of the early insights and possibilities since they hired Dr Stuart Morgan from the Australian Sports Commission and fully embraced it.

The AI journey has started via optical recognition from screens to give more accurate and timely data on player positioning, match-ups and interchange movement.

To explain its initial use, Kelsey takes CODE Sports into the middle of a Collingwood vs. GWS clash.

Normally Craig McRae’s lieutenants would be busily watching the who, where and when of superstar Nick Daicos’ opposition match up and relaying what they see as the defensive mechanisms being put in place to ‘tag’ him.

However AI will soon provide definitive data that supersedes the ‘eye test’ of assistant coaches.

“Instead of just ‘Nick Daicos is being tagged by Toby Bedford’, we will end up with a much more nuanced answer,” Kelsey says.

“The AI answer might find that he is tagging him in stoppages, but then when they’re outside of stoppages they’re not anywhere near each other.

“So you can get much more fine-grained and more accurate information.”

AI could soon help Nick Daicos break the defensive tag of players like Toby Bedford. Picture: Michael Willson/AFL Photos via Getty Images

Over in Adam Kingsley’s coaches box he may be frustrated that Darcy Moore seems to be unaccountable for a player and intercepting everything when the Giants go forward.

Optical recognition AI will be able to tell Kingsley exactly where Moore is setting up and running every time the Giants gain the ball.

“If Collingwood are going with a plus one, where’s the plus one, what are they trying to do and how do you counteract it,” Kelsey says.

“So we’re looking at what players are doing when they’re not around the ball or not applying pressure. Are they defending? Are they trying to recover? What’s their contribution when they’re not actually in an event.”

At the moment this new information is not quite coming in ‘real time’, though the more advanced systems being used in the NBA only take about 10 seconds.

Kelsey expects the time lag for AFL will be close to that soon, meaning the likes of McRae and Kingsley will not have to wait until Monday to see how Moore beat them or how Daicos was quelled. They’ll be able to try a change before it’s too late.

So how will this now be possible?

At the moment Champion Data uses a manual system where one person calls what is happening on the field and another keys that into the system.

By now using AI optical tracking on the screen, many things in that process will be automated.

“As opposed to just trying to do it with humans, we can start to work out things like who’s coming on and off the ground,” Kelsey says.

“In the change movements we can look at player positions, who’s playing on who and get them much more accurately.

“And then once we’ve got those sorted, we’d then start to look at some of the harder things like trying to work out whether we can detect kicks and pressure.”

Younger coaches like Sam Mitchell may be more willing to have faith in AI-generated data. Picture: Steve Bell/AFL Photos/via Getty Images

Just how much AI is used in the AFL in the future is only limited by our own perceptions.

“The major challenge with all this stuff is human’s acceptance of it,” Kelsey says.

“So if you think about driverless cars, all the research information says that a driverless car is safer than the human behind the car, but no one’s really prepared to take the leap.”

The consequences are not life and death for a coach, but trust in AI over gut instinct informed by a life in their sport has similar barriers when it comes to career longevity.

“Younger coaches that are more comfortable with that sort of technology might work out ways to use it and give them an advantage, but I imagine it would be a journey.”

Which is not to say Kelsey doesn’t think it would work. It could even mean revisiting Hopkins’ prediction that a computer could coach an AFL team

“I think in time that debate will be whether a computer using artificial intelligence might get a better result than a coach,” he says.

“But that’s way down the track.”

Originally published as Expert reveals how AI is set to transform the AFL and change how game is coached