The Science of Moving Dots

A guest post by Allison Kotleba:

When most people think of basketball, they picture the tall players, the fast-paced plays, and the seemingly impossible shooting skills. However, spatiotemporal pattern recognition does not Allison Kotleba_925519_assignsubmission_file_Kotleba Blog Post Ted Talks Picturecome to most people’s minds when discussing the game. In his Ted Talk titled The Math Behind Basketball’s Wildest Moves, Rajiv Maheswaran discusses the use of spatiotemporal pattern recognition in analyzing the players’ movements and using this analysis to help coaches and players create effective game strategies. This up-and-coming science aims to understand and to find patterns, meaning, and insight in all of the movement in our world today.

What is spatiotemporal pattern recognition? In layman terms, it is the analyzation of “moving dots.” For those more interested in the technical aspects behind this pattern recognition, the University of North Carolina released an analysis of a number of papers that are good examples of spatial-temporal modeling, a source which is readily available online. This very new kind of data is becoming more and more popular, especially with the popularization of devices such as cellphones and GPS. Because of its newness, data scientists have the challenge of finding patterns within the data. In an interview with writer Ben Lorica, Maheswaran explained these difficulties, “There’s no language of moving dots, at least not that computers understand…Allison Kotleba_925519_assignsubmission_file_Kotleba Blog Post Pattern GraphThere’s no computational language of moving dots that are interacting. We wanted to build that up.”

Here is one example of how this data can be used. To the left is a bubble chart with each bubble representing an NBA player. On the X-axis is their shot probability, and on the Y-axis is their shooting ability. If you take a player who generally made 47% of his shots, before, that was all you knew about him. Now, scientists can tell he would take shots that an average NBA player would make 49 % of the time (shooting probability), and they are 2% worse at their shots (shooting ability). This is significant for teams because it allows scouts to distinguish between all of the 47% shooters and determine their relative shooting ability and probability to the other 47% shooters.

What does this new science have to do with sports such as basketball? Maheswaran explained to Lorica that sports is one of the areas in which there is really great data available. Maheswaran said that “in sports in the last year, there have been tracking technologies placed in all major sports where they’re tracking all the players and the ball at a very, very high frame rate.” This availability of data as well as the large amount of people interested in finding patterns in this data, such as coaches and front offices, makes sports one of the best places to start building this science.

An important tool in developing this science is the use of machine learning, which allows scientists to go beyond their own ability to describe the things that they know. By giving the machine specific examples of movement and specific examples of non-movement, these scientists can teach the machine to see the game through the eyes of a coach. The machine is able to find features that enable it to separate particular movements and to discover the relationships between these movements. With this new information, new game strategies are being formed that are helping teams win games. In the near future, Maheswaran believes that real-time data will not only become a game changer, but also will help us to move better, move smarter, and move forward.

Maheswaran is also the CEO and founder of Second Spectrum, a company that applies analytics to sports tracking data. Those at Second Spectrum specialize in creating products that “fuse cutting-edge design with spatiotemporal pattern recognition, machine learning, and computer vision to enable the next generation of sports insights and experiences.” The company’s main goal is to revolutionize the way that people play, coach, and watch sports.

As Maheswaran mentions in his TED Talk, spatiotemporal pattern recognition can be used for much more than just sports analyzation. Writer Maureen Dowd fears other ways in which this type of data tracking can be used. In her New York Times article titled “Walk This Way”, Dowd discusses the Pentagon’s attempt at creating a grand database that can be used to track Americans’ every move, both literally and virtually. The Pentagon has been developing this technology as an antiterrorist surveillance system. The report outlining this research and development states that the “goal of this program is to identify humans as unique individuals (not necessarily by name) at a distance, at any time of the day or night, during all weather conditions, with noncooperative subjects, possibly disguised.” Though the reasoning for this data collection is for the good of the general public, many people like Dowd may find it to be an invasion of their privacy. Despite this, most, like Maheswaran, still have hope that this science will help to revolutionize how we think about movement in our world today.

Works Cited:
Dowd, Maureen. “Walk This Way.” New York Times 21 May 2003. National Newspapers Expanded. Web. 31 March 2016.
Lorica, Ben. “The science of moving dots: the O’Reilly Data Show Podcast.” O’Reilly. O’Reilly, 20 November 2014. Web. 31 March 2016.
Maheswaran, Rajiv. “The Math Behind Basketball’s Wildest Moves.” TED 2015. March 2015. TED. Web. 30 March 2016.
Revolutionize Sports Through Intelligence. Second Spectrum. Web. 31 March 2016.
“Spatial-Temporal Models.” University of North Carolina, University of North Carolina, n.d. Web. 24 April 2016.

About Sarah K. Salmon

I am a graduate student in mathematics studying algebraic combinatorics flavored by Coxeter groups at University of Colorado, Boulder. I earned my B.S. in mathematics at Northern Arizona University in May 2014.
This entry was posted in Math, Math in Pop Culture and tagged , , . Bookmark the permalink.

2 Responses to The Science of Moving Dots

  1. Daniel John says:

    Nice Blog. I have gone through similar blog few days back explaining the relationship of a data science and Golden State Worriers. They showed a graph which can explain every move of a rival team which actually was a successful approach with the data science. If you are interested you can read the referral blog here:

  2. Vimell says:

    Is there a link for the papers released by University of North Carolina on Spatial-Temporal Models? Thanks for your help.

Comments are closed.