Sports analytics blocks out human bias and picks up patterns and correlations to beat the competition.
In sports, data scores big as a team player. In fact, other than finance, no other industry competes with the intensity of sports fans, coaches, and players pouring through data. Their precision stat tracking and appreciation of quality data is so high that sports analytics is a fertile field of dreams for machine learning.
With so much historical and real-time data available in structured and nonstructured formats, coaches are calling data’s number to make decisions that were once based on gut instinct and experience. Machine learning is responding with consistent head-turning performances.
When machine learning weighs in against human decision making, the advantage goes to machine learning. Machine learning sidesteps human biases. It also has a much higher capacity of processing the vast data available and finding patterns and correlations in extremely large data sets. What further gives machine learning an edge, is that it’s always learning and improving. It knows the players, the competition, and the odds better than anyone—and its predictions and forecasts get better with time.
Machine learning is poised to be a game changer across all sports. These real-world highlight reels are worth studying to understand why machine learning is good for teams and fans.
Machine Learning Calls the Plays
German tennis star Angelique Kerber first picked up a racket when she was 3 years old. Two decades and thousands of backhand shots later she’s netted 10 World Tennis Association titles and a Silver medal in the 2016 Olympic Games. Certainly Kerber is physically gifted, but she also trusts predictive analytics and machine learning to beat her opponents.
She and her coach analyze her performance pre- and post-match, and they discuss the best tactics to win the match. To build her strategy, Kerber enlists the help of machine learning which analyzes every hit, mishit, spot where the hit bounced on the court, speed of the ball, wind velocity, and court temperatures. Her coach can identify her competitor’s weaknesses by asking about forehand vs. backhand mishits, even drilling down to serve returns.
During matches, Kerber’s coach can query the data stored in the cloud in real time to find vulnerabilities and advantages to help her score points. In 2017 Kerber climbed her way to a No. 1 ranking, proving that applying machine learning can be as powerful as a wicked cross court shot.
Score One for the Fans
Raw numbers are common fare for fans but height, weight, free throw percentage, and runs per carry only tell so much. Facing major gaps about their athletes, coaches are turning to machine learning to better understand players’ stamina, conditioning, and performance. A number of NBA teams video player movements during practices. With 82 games in a regular season, players are more prone to injuries when they move into the red zones of fatigue and exhaustion. Coaches are referring to the data to support decisions for sitting players out for a rest to prevent injury. A combination of fatigue signs and proof of diminished capacity in the data are signals for a much-needed break from the action.
Collected and analyzed data includes player movements at 25 frames per second, heartrate, and movement patterns—stops and starts, bends and twists, and direction. The data provides a complete picture of how each player is moving and how his body is responding.
For now, the NBA has ruled that wearables are not allowed during games, but the data is impacting the game. Multiple teams have rested players during the fourth quarter or benched them for entire games believing that the rest will lead to improved future performances and prevent injuries. For fans, not seeing a favorite star play is disappointing, especially for big ticket games. Long term, though, if machine learning can keep players safe and in their uniforms rather than street clothes, that’s good for both the team and the fans.
DataGeniuses Avoid March Madness Bracket Busters
In every sport, there’s an element of luck. A game-changing bad call from a ref or a freak accident that causes a star player to leave the game both fall into the luck bucket. These events can upend expected outcomes and completely re-write a coaches strategy. Those events also remind us that analytics isn’t a perfect crystal ball.
In a unique turn, machine learning addresses the luck factor and applies it in algorithms in an attempt to understand the unknown and account for it. The SAP DataGenius team, for example, took on the 2017 March Madness bracket challenge and factored in luck. Overall, the goal was to do very little to the data and let the predictive analytics with machine learning software do the heavy lifting. Identifying the winners would require data preparation, modelling, data exploration, planning and what-if analysis, visual storytelling, and automated smart data discovery to uncover patterns among each team playing in the tournament.
The team started by collecting data from public websites to create a unique data model using SAP Analytics Cloud for their analysis. The next step was preparing the data; that task required only minor changes such as updating headers, adjusting columns and rows and removing messy data. Once cleansed, the team moved into exploration mode, asking questions of the data and reviewing visual charts to better understand the data. Next machine learning was put to work to uncover correlations and insights such as which team stats are indicators of success. At that point, it was time to add the luck percentage to calculate which teams have the best chances to pull off an upset.
After churning through the data, the software picked a winner for every matchup, and it accurately crowned North Carolina as the tournament champion. The team finished with a bracket of 46 wins out of 63 games, giving them a 73% correct rate.
As a testament to machine learning’s ability to consistently improve, this year was the first time in five attempts that the software predicted the winning team. The plan for next year is to improve the models for those hard-to-predict, evenly matched 7 and 8 seeds.
There’s an A in Team; and It’s for Analytics
When Michael Lewis’ 2003 best seller “Moneyball: The Art of Winning an Unfair Game” exposed how the Oakland A’s used data to beat teams that had much bigger budgets, it fueled the imaginations of fans, coaches and owners from every sport. The vast majority of teams now employ one or more data experts now, and these elites are as serious about honing their analytics skills as the athletes. While games are fun, winning is more fun and the team with the best data and data experts can provide the insights that will give the players a huge advantage over the competition.
Mike Flannagan, SAP
Mike Flannagan is Senior Vice President of SAP Analytics and serves as Global Head of Market Strategy for SAP Leonardo, SAP’s newly announced Digital Innovation System. SAP Leonardo provides accelerated digital transformation solutions by focusing on well understood industry problems and repeatable solutions that combing Analytics, Big Data, Machine Learning, Internet of Things (IoT), and Blockchain, built on the foundation of SAP Cloud Platform.