How to find new football stars with AI technology
For football (or soccer, for American) fans, talent on the field is a thing of beauty. Names like Messi, Ronaldo, Pelé and Zidane evoke images of unparalleled grace and athleticism. As we saw in this year’s World Cup, new stars are emerging all the time.
But have you ever wondered how new talent rises to the top? What if AI could help find the next football star?
Recognizing potential is a human talent
At JUST ADD AI, our mission is to unlock human talent. We’re using AI technology to automate repetitive and mundane tasks, so people can focus on what they do best: complex problem solving, critical thinking, creativity and artistic expression. We believe we can make work more productive, more rewarding and more fun.
We were excited to make this vision a reality in football, with a German Bundesliga team. If you pay attention to the sport, you know that the Bundesliga is one of the most competitive leagues on the planet. And lining up the right players is crucial for a winning team.
But there has always been an element of intuition in the process. Teams dispatch scouts to championship games, relying on their deep experience to spot fresh talent. Gut instinct is important here—there could be a player with great potential whose physical abilities haven’t fully developed, or one with strong physical prowess but a bad attitude.
Knowing how to parse these traits is a uniquely human talent. So how could AI possibly help?
Scoring insights in a field of unstructured data
When we worked with the scouts at this football team, we found that the talent acquisition process has its challenges.
This particular team was sitting on something like 100,000 scouting reports it had accumulated over the years. The information in these reports is gold, with insights into playing style, work ethic, strengths and weaknesses, team dynamics, etc.—all filtered through the expert eye of the talent scout.
The problem was that the team couldn’t really access the information in a useful way. Nobody wants to read thousands of reports just to check a hunch or find one particular player.
It was an ideal opportunity for us to introduce AI.
We built a new product called JAAI SCOUT, which uses IBM Watson Knowledge Studio and Watson Natural Language Understanding to extract insights from unstructured data and pull them all together into a single dashboard. We also use Watson Personality Insights and Watson Analytics for Social Media to provide a deeper perspective on players.
We actually trained Watson to understand scouting reports and extract the most relevant details, making them searchable and available for visualization. The reports are still the most important element—JAAI SCOUT just makes them more accessible— so scouts don’t feel they are being replaced. In fact, AI is making their expertise even more influential in the team’s decisions.
AI makes a difference you can see on the field
It was January 2018 when we first saw our AI solution in play. Our customer used JAAI SCOUT to make a decision on a certain player during a Bundesliga transfer period. In the end, it worked out very well for the team.
I’ll admit, I got goosebumps when the new player scored his first goal and the entire stadium started shouting his name. I still get goosebumps when I think about it—because, in a small way, I was part of that process.
Watson has been a great complement to our own deep learning technology. With pre-trained or custom-trained Watson instances, we can develop solutions very quickly, using the outcomes of those models to feed into our own neural networks. So, for example, we built neural networks to predict fair market values for football players, taking into account all the in-depth analysis on their characteristics.
By bringing all the data together into a cohesive picture for football teams, we’re providing a very powerful combination that has the potential to up-level the game. We can’t wait to see what’s next for the sport of football.
Listen to Roland Becker to find out how IBM Watson has made their business smarter: