In the runup to the NCAA's March Madness tourney, the guys in my house have been watching lots of college hoops. I usually don't join them, but even if I'm just listening to a game with half an ear, I'm sure to get an inundation of statistics-related commentary.
A lot of these stats are superficial, really -- basic stuff like number of games played, steals, and assists -- that any diehard fan could rattle off easily. Knowing the number of three-point shots attempted and the percentage made isn't going to get any team a competitive advantage.
Every sport has these sorts of basic insights on player and team performance, and many teams are applying them in the aim of picking the best players, putting together winning teams, and making sound decisions on the court (or, as I mentioned yesterday, on the field). These are "table stakes" in sports performance analytics, Tom Davenport writes in "Analytics in Sports:
The New Science of Winning," a February International Institute for Analytics research report sponsored by SAS (which also sponsors this site).
In his report, based on a series of interviews with professional sports teams and vendors in the US and Europe, Davenport identified seven types of table-stakes analytics: use of external data sources, descriptive analytics on players, optimal lineup analytics (basketball), player scoring for draft analysis, player salary optimization, simulation of games, and analysis of game tactics. Teams need strong players and good coaching to win, but these analytics "have certainly become established as important augmentation for those basic success factors."
But let's look ahead to the types of player and performance analytics that are not quite so common yet. Davenport called out five "frontier" analytics applications and said that only a few teams are using them aggressively today. Could these applications really open the playing field, so to speak?
1. Analytics on video
Davenport called video the frontier source of data across all sports. Major League Baseball (MLB) and the National Basketball Association (NBA) have standardized on video capture services, which eases the pain of tagging and editing the video content for analysis. Other leagues, such as the National Football League (NFL) and the National Hockey League, haven't standardized on a video vendor, so teams are left to address the tricky tasks of editing, tagging, and analyzing video on their own. Regardless, the level of analytics here varies, from the simple tracking of descriptive analytics like ball touches and rebounds in basketball to the highly complex analysis of, say, how often a particular player goes left when driving toward the basket from the free throw line.
2. Analytics on location/biometric data
This includes data from GPS devices, radio frequency devices, accelerometers, and other types of biometric sensors. While the location data allows you to assess total activity, like miles run or steps taken, biometric data gives information on physical activity. "It's also possible to use this type of data to understand interactions between players, but this will require greater sophistication in data analysis," wrote Davenport, who will address this topic, among others, during a session on big data in sports at the MIT Sloan Sports Analytics Conference tomorrow.
3. Gathering and using proprietary data on players
Player and team data is readily available from leagues or external vendors, but a "few highly analytical teams... gather their own proprietary data, or adopt technologies that produce it." These teams include the NBA's Houston Rockets and Orlando Magic; MLB's Boston Red Sox and San Francisco Giants; and the NFL's New England Patriots.
4. Engaging players in analytics
A lot of sports experts, coaches among them, see no value in sharing analytics with players about how they or their opponents perform. However, showing players the data shouldn't be ruled out in all cases. Davenport cited some notable examples, including Patriots quarterback Tom Brady and Brandon McCarthy, a pitcher with MLB's Arizona Diamondbacks.
5. Open data analysis by fans
What's new here isn't the use of fan data, but the actual sanctioned use of such data or even the hiring of fans by teams. As one extreme example, Davenport cited Manchester City of soccer's English Premier League. "That organization has partnered with its primary data provider Opta to make all player performance data from all EPL clubs available for analysis by fans and researchers." To date, 5,000 fans have downloaded the data.
That last frontier is one of the most intriguing to me, pointing as it does to the sheer volume of data being collected on individual and team performance. Who wouldn't want a little help in analyzing it, especially if management isn't on the same page as coaches with their understanding of and desire for analytics? But which of these frontier analytics applications strikes you as the most interesting? Do you think any might provide a competitive advantage? Share below.
— Beth Schultz, , Editor in Chief, AllAnalytics.com