Fox Sports Australia predicts wickets with AutoML Tables

Fox Sports Australia has rolled out a machine learning model powered by Google Cloud’s AutoML Tables service to predict when wickets might fall in live cricket matches.

Dubbed Monty, the model was trained on a year’s worth of historical data on cricket matches licensed from Opta Sports to predict when a wicket could fall five minutes before it did, according to Jack Smyth, head of innovation at Mindshare, the media agency spearheading the initiative.

Smyth said the data was in structured form, and was typically used to provide score updates or fan trivia, but “the power of AutoML Tables allowed us to look at that data to understand not only when a wicket would fall, but how”.

“This delivers incredible insights to fans on when they should watch and what could happen next,” Smyth told the media on the sidelines of Google Cloud Next’19 in San Francisco.

Monty was designed to parse 83 variables within seconds after a ball is hit. During this prediction window, it also factors in data on recent shots in a game, so that it understands what is happening on the pitch to make its predictions.

Smyth said the insights are then displayed on electronic billboards across major Australian cities, and as notifications in the Fox Sports cricket app, which delivers commentary on the fly. The predictions are even used by Mindshare to determine how much it pays for Google advertisement impressions in its promotional campaign.

“We are also making ads on the fly based on what could happen next,” said Smyth, adding that cricket fans could even ask about Monty’s call on a wicket, thanks to the predictive power of AutoML Tables and the personalisation capability of Google Assistant.

“People can ask how the model makes its predictions and have their personalised commentator through Google Home and on their mobile – that’s our vision of the future of live sports where you are in control of what you’re watching, and what you should focus your attention on and why,” he added.

As a result of its Monty initiative, Fox Sports Australia has increased user engagement by 140% while delivering 150% more subscribers per dollar spent by communicating to cricket fans in the right place at the right time.

“Given the success we’ve seen, Google has set itself apart not only in the strength of its team back in Sydney, but also in its ability to provide native interactions and integrations with a broader suite of services,” said Smyth.

“As a brand, that gives us unparalleled ability to manage our campaign with a single source of intelligence across every single touch point.”

Now available in beta, AutoML Tables enables organisations to build and deploy machine learning models on structured datasets with no code.

With just a few clicks, data can be ingested from BigQuery and other Google Cloud storage services into AutoML Tables to build and deploy machine learning models in days rather than weeks.

Industry observers have billed AutoML Tables as a notable addition to Google Cloud’s suite of AI services – by making it easier for organisations to dip their toes into machine learning using structured data that a majority of them are already managing.