The Blog


At an IBM Developer Outreach event in Lower Manhattan in late August, curious developers were treated to a couple of fascinating talks about how IBM is using AI to enhance sports fandom. First up was data scientist Aaron Baughman, who explained how he was able to sweep a fantasy football league with a little help from Watson. Then IX’s lead solution architect David Provan shared about how his team’s work has enhanced the experience of attending one of tennis’s biggest events.

How to Play Moneyball with Watson

In the 2011 baseball film “Moneyball,” Oakland A’s general manager Billy Beane says the following:

“It’s about getting things down to one number. Using stats to reread them, we’ll find the value of players that nobody else can see. People are overlooked for a variety of biased reasons and perceived flaws. Age, appearance, personality … mathematics cuts straight through that.”

The film tells the story of a real-life revolution in how baseball managers made decisions based not on intuition but hard data. Beane found a number-crunching Yale grad to help him focus on buying “wins” rather than players, and to that end, they assembled a team using by statistical analysis. This enabled them to secure players which were high value in proportion to their low cost on the market.

When listening to IBM’s Baughman speak on the subject of fantasy football, it’s hard to not feel like you’re witnessing another revolution in sports analysis. He opens his talk with a startling boast: He used Watson™ to dust the competition: 13-0 in a 14-team league.

Photo of IBM Developer Outreach event

Baughman explains that most people use an average of 3.9 sources of data to make decisions in fantasy football. Some play around with spreadsheets and develop sophisticated tools to track and analyze player performance. But no matter how hard they try, they can never shake their human bias.

Baughman’s team taught Watson to analyze thousands of unstructured data sources (roughly 1TB of data) in media – articles, podcasts, video commentary. This unstructured data is not normally factored in fantasy footballers’ spreadsheets, but it is a goldmine for Watson. Baughman used natural language processing to evaluate media sentiment toward each player and team, and translated the combination of raw player performance data (e.g., injury reports and traditional sports statistics sources) and media sentiment into numbers, fed that data into deep neural networks, and used the resulting information to draw conclusions about how to best manage the team. They eliminated bias by casting a wide net for data sources. For example, local media might contain more data about an obscure player, but national data might be less likely to possess bias in favor of the hometown player. Once Watson had crunched the data on each player, the platform assigned a “boom” or a “bust” percentage, as well as an injury probability. Lastly, Watson assigned point projections for each player, factoring in age, height, and other data, which was used to run simulations and spit out a point spread. This enabled Baughman to compare players before every game.

This technology is obviously bigger than fantasy. The game has 60 million fans in the United States, but the innovations being pioneered by Baughman and other data scientists at IBM promise to influence every sport, to say nothing of the wider worlds of business, medicine, law – truly every human endeavor that involves complex decisions based on lots of data.

Elevating the experience at the U.S. Open

The United States Tennis Association is known for holding the US Open, but the organization works year round to grow the game of tennis. IBM’s David Provan and his team work with the USTA to provide tennis fans with experiences that will enhance their experience at the big annual event.

This starts with the US Open App and website, which includes a data analytics platform and a Watson-enabled virtual assistant, bringing data on players, scores, related news, and helpful tips to tennis fans throughout the games in real-time via the IBM Cloud. In addition to the US Open App, the Virtual Concierge is now available on and optimized for Facebook Messenger. The USTA provisioned six Cloud Foundry services to implement the solution – three in the United States and three in the United Kingdom, allowing for load-balancing between the regions.

The Virtual Concierge learns as it interacts with users. Users can ask questions like “What’s the weather like at Arthur Ashe stadium?” or “Who won the latest women’s match?” or “Where should I get dinner before the match?” The Virtual Concierge can also leverage map data to direct users to points of interest near the venue. If the Virtual Concierge is unable to arrive at a helpful answer, a human can step in to help with live chat. It’s stateful, meaning that it remembers previous conversations with users and can build onto those past experiences.

IBM also brings AI-powered highlights to tennis fans. The technology can recognize signals from game footage like crowd noise, match scoring, and player gestures. For example, if the crowd goes wild after an ace, and the player who performed it has a big toothy grin on her face while pumping her fist in triumph, AI highlights interprets that behavior as being indicative of a highlight-worthy moment. It uses these data points to build an “excitement score,” then feeds the top highlight videos to the fans.

Both talks were fascinating looks at the way technology is enhancing the experience of sports for fans, and both enabled developers to learn how AI is being used in tangible, fun ways. Read more about how an IBM team is using AI to make ESPN Fantasy Football predictions.