AI In Sports Project = Performance Analysis + Performance Control, Where to Start

 

Data and the business of sport have become inextricably linked. We may not hear the expression ‘big data’ used so often these days but that doesn’t mean there’s not a huge amount of it about. So much so, that many sports executives are finding themselves having to grapple with artificial intelligence, often without the first idea of what an AI project really should look like.

Like many technologies in sport, AI has an extensive list of jargon associated with it, which has contributed to the sense that it is an impenetrable dark art which is probably beyond the budgets of many sports. This has undoubtedly prevented sports bodies from embracing it.

So where can an organisation that is data-rich and AI knowledge-poor start?

At a recent session of the STA Group’s ‘Access Innovation’ sports technology coaching programme, AI experts attempted to demystify the terminology.

The panel included Catherine Breslin, the founder of Kingfisher Labs AI consultancy who previously helped create Amazon Alexa; Max Métral, senior analytics manager at Formula 1; Lucas Galan, head of applied data science at specialist AI company Flamingo; and Jonathan Boase, founder of Koroibos Ventures and formerly IBM’s senior European lead for Cognitive Analytics in Media and Entertainment.

What is AI?

‘Data and analytics’ is a phrase that widely used in sport but what does it actually mean and how does it translate from information harvesting to useable intelligence?

Breslin explained that “Artificial Intelligence is a phrase that dates back to 1956 and most commonly means machine learning. There is a raft of additional terminology (deep learning, reinforcement learning, unsupervised learning) but it all means training software to process data in all its forms (images, numbers, words, sentiment) and establish patterns or answers.

“Where information is nuanced, the output can be flawed. Social media searches are a good example of this – if you search ‘Baltimore Ravens’ you are as likely to get images of birds as anything pertaining to the NFL. However, once the interrogation process is robust, the opportunities to apply AI across a wide range of business interests are huge.”

Where to start with an AI project?

Sports have been gathering data of all types for many years, often doing so because of a sense that they should collect it, rather than having any clearly defined purpose of what they plan to do with it. How then, does an organisation take its pool of data and apply AI tech to achieve a given outcome?

Galan, veteran of AI projects for M&S, GSK and Diageo, said: “Crucial to a successful AI project is understanding what problem you are trying to solve and work backwards from there. Embarking on something with a clear idea of its intended conclusion sounds obvious, but failure in this first step is probably the most commonly-made mistake.”

For Formula 1’s Métral, the innate culture behind a project is also crucial: “Whilst AI is essentially data, you don’t have to be a data scientist to ensure a successful AI delivery. For an organisation to embrace AI adoption successfully, there needs to be a data-savvy intermediary in-house, backed by top-down support for the project,” he said.

“Clear and transparent communication is fundamental to the success of data-led projects, so this should be a person who understands data within the organisation and how to access it. They also need to translate the organisation’s needs and objectives to the AI experts, then convey the AI team’s progress to their wider colleague network. Also, AI isn’t just about expensive solutions, there are free tools available that do a perfectly good job and create value.”

Boase also had some reassurance for smaller organisations: “I frequently work with businesses who think they have nothing in terms of data, experts or resource but you can create infrastructure and find data,” he said.

“As with all data, rubbish in means rubbish out, so a wise starting point for any project is to establish data quality and any gaps between what you need and what you have. The good news is that, with the advent of synthetic data, poor or ‘gappy’ data has never been easier to fix; synthetic data is an emerging discipline that can accurately recreate required information.”

Getting past human instinct

Another part of creating the right business culture to support AI work is ensuring the organisation is open to what AI-led findings are telling it and the methods taken to achieving that.

Galan explained: “Heuristics – or short-term human decision-making – needs to be overcome as this is what we are comfortable defaulting to. For example, we have all seen senior executives who like to scroll through fans’ comments on social media and draw conclusions based on that but unless they can process many thousands of comments, quickly and accurately, human limitations and interpretations will skew the data. Processing huge data quickly is where AI can really add value to your work.”

Realistic expectations and the power of pilots

Having committed to applying AI within an organisation, the panel emphasised the importance of setting realistic expectations.

Boase explained “People need to understand that test pilot schemes and the attitude of ‘good enough’ are crucial to setting expectations amongst colleagues.

“‘Good enough’ when seeking results might sound like a low bar but it is about being realistic; you don’t necessarily know what you will find in or from the data when you start out, so over-promising and under-delivering is ill-advised. In regard to pilots, I’ve yet to see one fail but even if they do, the mindset should be ‘it was only a pilot’.”

Métral added: “In starting a project, don’t feel that you have to go it alone. Sport can be a collaborative industry so reach out to your peers and talk to experts to find out what their experiences have been.

“Also, put a brief out to different suppliers and listen to what they have to say – and avoid anyone who makes AI sound inaccessible. A project should be a transparent process and agencies who suggest otherwise are best avoided.”

Rebecca Hopkins is the chief executive of The STA Group and moderated the Access Innovation panel. SportBusiness is a media partner of the STA Group and the annual Sports Technology Awards.

This post first Appeared on SportsBusiness.com