IBM’s Watson proves AI can understand the offside rule

Conversational AI powered by natural language processing is making sports analytics more accessible – at least to those who can afford it

As glorious a venue as it is, Twickenham, the home of English rugby, seemed an odd place to showcase how artificial intelligence (AI) is disrupting football.

Nevertheless, IBM invited analysts and reporters from the worlds of sports and technology to the stadium for an event hosted with semi-professional football side Leatherhead, to explain how IBM’s AI assistant Watson had assisted the side’s management during this season’s campaign.

Created in collaboration with IBM Master Inventor Joe Pavitt and Leatherhead’s management team Nikki Bull and Martin McCarthy, the tool uses natural language processing (NLP) to create scouting reports, based on unstructured data from Twitter feeds and online match reports, that illuminate opponents strengths and weaknesses to inform tactics and selection.

“Using Watson, we can produce reports that can tell you if a player was presented in a positive or negative light in relation to short passing,” Levitt explained.

It also uses NLP to enable players and staff with limited technical aptitude to easily recall video footage of match events and relevant analytics, including player transition and formation effectiveness. This required Levitt to teach Watson football terminology such as “4-4-2”, and “offside.”

“Watson sits as a layer between to allow you to have a conversation with that data and with that video,” he said.

The great equal-AI-ser?

Leatherhead — who were battling relegation when IBM first approached them about the possibility of creating an AI tool — rose the table over the course of the campaign, eventually finishing eighth and at one stage having a decent shout of promotion.

McCarthy was reluctant to credit IBM Watson as playing a vital role in Leatherhead’s success, but did say the tool helped his cash-strapped team gain an edge, without having to invest in a sports analysis team (typically the preserve of elite sporting outfits).

Although AI is often decried for taking the humanity out of society, “the edge” McCarthy praised was the way in which the tool enabled management to be honest with players about where their performances were lacking.

“Unfortunately players might not respond well to criticism, but when me and Nikki explained to them that ‘it’s come from this [IBM’s tool]’, that’s where it really came in handy,” he said. “At the end of the season the players clapped us because of the relationships we formed, and Watson was a big factor in helping us form those relationships.”

Other Watson applications have been criticised for requiring too much upfront investment and staff training to be practical for the average enterprise. As this was more a proof of concept/marketing tool to demonstrate Watson’s clout, Leatherhead didn’t have to concern themselves with the financial side of things.

However, the data analysis they were recalling depended on a sophisticated supply of data that IBM sourced in partnership with sports data juggernaut Opta, that was later integrated by the Levitt. Joe said that two Opta staff were required to attend every game.

Whether teams of Leatherhead’s limited wealth could afford to purchase this data supply and hire data staff to integrate it questionable. But there is no reason teams higher up the food-chain won’t soon look to invest in Watson to replicate Leatherhead’s success.

The NLP event and analytics recall could prove particularly useful at elite levels, where there exists an abundance of analysis, precious little time to analyse it, and few who have the skills. IBM’s tool makes the process efficient and accessible: players can use everyday football terminology to recall intelligent analytics, and as the tool is hosted on IBM’s cloud they can take it home too.

“You can certainly see the players that have used it more, and those who haven’t,” said McCarthy.