Blogs

Using predictive analytics to win

Post Comment
Director of Interactive Sales, WNEP-TV / WNEP.com

It’s that time of the year again, when fans of American football will be lining up at the supermarkets to plan festivities for Sunday's big game. Families will gather around a living room television and friends will collect at their favorite watering hole to watch the New England Patriots take on the Seattle Seahawks for the league championship. Everyone watching the game will have an opinion on who they think will win. Casual observers may pick their favorite based on the color of a team’s uniform. NFL aficionados will have years’ worth of stats memorized in order to argue why they feel their team will triumph. Las Vegas will set their betting lines based on how best to maximize wagering on both sides of the matchup. Everyone will argue for their favorite way of picking a winner, but I’m left to wonder: is there a better way?

No matter what type of observer you are, you simply cannot argue that any number of infinite factors will affect the outcome of Sunday’s game. Factors such as temperature and barometric pressure affect the coefficient of friction on the ground, balls and players uniforms. Things like injuries, the amount of time spent in travel and personal factors will affect the players’ individual performance. The amount of sound generated by the fans in a stadium and the acoustics determining how that sound reverberates will govern how easily players can hear the quarterback’s calls. At this point, I’m sure you are beginning to understand the incredible level of data that would be required to accurately model and predict, with relative certainty, any single aspect of the game.

Enter predictive analytics

In a previous article, I wrote about how small to mid-sized businesses can take advantage of intelligent predictive analytics with a very minimal barrier to entry, and a business the size of the NFL should be no different. Intelligent predictive analytics tools, such as IBM’s Watson Analytics technologies, are now providing ways in which we are able to discern patterns which may be found in vast quantities of data, previously locked away in unstructured formats. Unstructured data like temperature, sound levels and the human condition now may be correlated to find predictive patterns intelligently suggesting outcomes. How can predictive analytics be used in analyzing the upcoming game, you might ask? Let’s take a look.

Predictive analytics and sports

Using predictive analytics in sports, to a certain extent, is nothing new. Baseball has undergone a radical transformation in the way in which it recruits and manages players, based on a field of statistical study known as Sabermetrics. Using Sabermetrics, teams are able to base managerial decisions on empirical data rather than solely relying on the “trained eye” of seasoned scouts and veterans.

Football, however, has largely been unsuccessful in embracing this type of performance governing mathematical forecasting. This is due to the relative complexity of the action on a football field, as compared to baseball. In baseball, play analysis may be more easily performed as it can be broken down easily in terms of a pitcher versus a batter. In football, you’ve got more than 20 moving parts on a field, any of which can easily affect the outcome of a particular play.  Multiply this by environmental factors, and you’ve got a nearly insurmountable task—or is it?

IBM Watson Analytics: The new sideline coach

IBM Watson Analytics has broken new ground in its ability to digest large amounts of unstructured data, and make intelligent predictions based on the patterns it observes. The medical industry is successfully using IBM’s Watson Analytics to treat complex cases in the field of oncology, and it is also being used to help streamline the process of bringing new pharmaceuticals to market by navigating the FDA’s laborious approval process.

Watson Analytics, and other predictive analytics tools, have the innate ability to not only process this data, but make repeated accurate assessments of future outcomes, based on past performance information. The sheer amount of data, which I mention above, governing outcomes in a football game would only help predictive analytics tools to further enhance the accuracy of their models. “Resolution of Data” is a term I like to use when attempting to describe the amount and quality of data that’s available for analysis and processing. Similar to a photograph that has an extremely high resolution, a dataset with an exceedingly high resolution would effectively reduce the instance of error when predicting the outcome of plays, or a full game.

For example, it’s said that onside kicks are successful over 60 percent of the time in the first three quarters of a football game. By taking this into account, in combination with historical performance data demonstrated by the individual players on a field, as well as current environmental factors, tools like IBM’s Watson Analytics may be able to effectively act as a trusted sideline assistant coach. Think about it: you are a coach whose team is competing for a national championship and, like it or not, your emotions will certainly play a role in the plays that you call. Emotions cloud judgment, and are capable of instilling a sense of self-doubt in tense situations. Predictive analytics sideline tools would function as that unemotional observer with suggestions that would become more and more trusted over time. Much like the oncologist who trusts IBM’s Watson Analytics technology to recommend a course of treatment for a patient with lung cancer, so too could Bill Belichick one day trust intelligent predictive analytics to make a call for an onside kick. 

Future applications could also be tied to integration with streaming real-time data, produced by in-uniform biometric scanners. These devices could measure physical statistics such as:

  •      The force of impact produced by a hit on various parts of a player’s body
  •      Vital statistics tracking real-time fitness and performance data
  •      Neurological scanners tracking mental activity and capacity

This data could be ingested into a real-time player management console, which would essentially remove the obstacle of a player performing at reduced levels due to an undisclosed injury. This would also assist in protecting a player’s health, in a league plagued by severe, career-ending injuries. Again, these decisions could be made impartially, by a predictive analytics engine trained to look for patterns in empirical data, and removing the emotion of a decision.

Where else could Watson Analytics be used?

Another major area of use could be in the officiating of a game. By combining visual recognition and motion capture technology, referees could be assisted in making game-altering calls by reinforcing their observations with statistical data. This could help to reduce the chance for error as well as the likelihood a call may be successfully challenged.

Predictive analytics are already being used by municipalities around the world to help reduce traffic and congestion on public roadways. So too could major sporting events not only use predictive analytics to reduce the time it takes to get into and out of various parking facilities, they could also help to govern the construction of new stadiums to better enhance efficiency.

NFL analytics in action

No one could argue that intelligent predictive analytics are playing an increasingly important role in our daily lives, and this will only become more obvious when they are successfully deployed in spectacularly public fashion. While we aren’t at the point, yet, of using these tools in real-time during live games, you could still get a taste of what that might entail. 

Currently, IBM’s Watson Analytics is pre-loaded with a free NFL dataset pertaining to 2014 offensive statistics. Right now, you can log into a free account and see for yourself just how flexible a tool like Watson Analytics currently is in analyzing data and providing meaningful insights in data such as this. This sample dataset provides statistics on items such as:

  •      Receiving Targets
  •      Receiving Yards
  •      Field Type
  •      Pass Completions
  •      Passing Touchdowns
  •      Stadium Type

Heck, maybe even Bill Belichick and Pete Carroll are already signed up!