Name: Brainstorm AI London 2024: How AI Is Enhancing The Scouting Process And The Fan Experience
Uploaded: 2024-04-15
Duration: 5 min 15 s
Description: Presenter: Dr. Ryan Beal, Co-founder and CEO, Sentient Sports

Presenter: Dr. Ryan Beal, Co-founder and CEO, Sentient Sports 

Brainstorm AI London 2024: How AI Is Enhancing The Scouting  Process And The Fan Experience

Photography by Joe Maher Fortune Brainstorm AI

Conferences

I'm Ryan Bill. I'm CEO and co founder of sent in sports where we help professional sports teams and organizations um utilize cutting edge A I technologies. So initially, we focused on scouting in football and helping scouts in 2020 who couldn't go to games find a new way to assess players. So at the time, a lot of the analytics in the space was focused on backwards looking data. So it's how players are performing at their current clubs and with their current teammates. However, in football, when a team, when a player moves all of this changes, so the player moves into a new team um into his new teammates. So you can see an example here for Harry Kane, um for his Bayern Munich move in the summer. Um He moved into the new team, had new teammates and then we could simulate a number of different things. So the first of these was team chemistry and this effectively looked at how well he would work with his new teammates based on things like his languages. He speaks how good he is as a player. And if he's played with any of these players before we then had tactical fit, which is a prediction of how well he'll fit in with the manager's tactics and whether or not he will suit the formation and style of the manager and then finally return of investment, which was a prediction of the suppliers like to sell for a profit in the future. So what this gave to decision makers was forward looking risk assurance effectively to help them make multimillion pound decisions. And the key thing we learned while working with some of the top clubs around Europe on this was around explainability and for a decision maker to be able to trust these models, they really had to understand what underpinned them and what drove those decisions. So you can see we were able to unpick some of these and provide different ways of doing it. So what we found with generative A I over the last few years as it's come to the mainstream is that it's able to solve different problems of sports. And one of these in particular is how they can start to generate new revenue from their fan bases. So at the moment, sports revenues are generated very much from physical sources, things like their ticketing, their merchandizing as well as TV, rights and broadcasting. Whereas in the future, we expect this to shift and it's be very much focused on digital experiences and the new ways that they can interact with their global fan bases. So we have built a system to be able to test this, the system is called scout G BT. Um And we'll show an example of this now and it's able to interact with all of our back end analytics. It is able to access all of the things we just showed and all of the data that we've got on football. So in the example we see here, um we've got an Arsenal fan who's asking questions of whether or not they think the arsenal will win against Bayern Munich. This week, the system is then able to go out and access all of the data we've got and simulate whether or not arsenal will win in that game. The fan can then ask what position should ask and look to strengthen in the summer and the system is then able to access all of the data we've got on their squad depth. So it will provide this back to the fan as well as providing a assessment and a chart of their squad. So it's able to do all of this in real time with real time data as well. Then finally, we can ask, you are the top five left back to ask Shahine to solve these, these weaknesses that we identified. And then through this, we can provide that back to the fan um as you can see here. So this is what's known as a Rag LM, which is a retrieval augmented generation system. And this allows us to provide an interface to all of that back end data that we've got, which allows the fan to access a domain specific and real time LLM, which provides a difference to things like chat GPT and anthropic exist at the moment, which ultimately are all trained on the same things like the open web. Whereas this can give it access to much more domain specific areas. So the more that a fan uses this system, the more that we can start to learn about what the fan likes, what their preferences are and ultimately, that can lead to ultra personalized content generation. So in the example here, we've learned that the fan supports Arsenal and we've learned that they are interested in transfers. So using that information, we can serve them with this article here, which is a simulation of Alfonso Davis potentially joining Arsenal in the summer. And then this is displayed onto the screen to the user. Now, if the user likes this article, they're able to ask questions to a virtual journalist so that we can then provide articles on the fly to their own preferences. So you can see here, they've asked for the same article, but for mbappe move to Real Madrid and then that's been produced and served back to the user. This is also supplemented with things like charts and images which are all displayed on the flight and eventually be multi modal with things like podcasts where if we know a user is at a train station and has half an hour to burn, we can serve them with an automated personalized podcast. This also leads into multi languages. So if we know that a user is in Thailand and they speak different languages to what we might expect where the clothes was based, then we're able to serve them with content that is unique to their interests, like the club and their players as well as unique to their consumption preferences in terms of length. And then also in their local language, which can allow sports teams to provide a much deeper connection to their global digital fan bases and ultimately generate more revenue. Thank you for listening and I'm happy to answer any questions later today.