Five Real-World Applications for Sports Analytics

When Moneyball first published nearly two decades ago, it was a chronicle of an industry slow to adopt advanced statistical methods for evaluating player worth. Since then however, and due in no small part to the book itself, sports is now at the forefront of applying advanced analytics to change how the game is understood by front offices, played by athletes, and enjoyed by fans. Advancements in cloud computing and AI have expanded the universe of inputs and methods of evaluation such that we went from basic box score statistics to now highly unstructured data like player, teammate, and opposing team position, movement, field conditions, weather, and exhaustion.

And by the very zero-sum nature of sports competition, it makes sense that competitors quickly imitate winning tactics. In each contest there can only be one winner, and winning can add millions to an individual athlete’s lifetime earnings, tens of millions to a team’s annual broadcasting rights, and hundreds of millions or even billions to the value of the team were an owner interested in selling. Normally success can be fleeting as rosters age or competitors quickly adapt winning tactics, but sports organizations have found that data and analytics can bring a long-term competitive advantage. For example, in 2015, star free agent LaMarcus Aldridge caused a stir by not choosing the Los Angeles Lakers due in part to his perception that at the time they were not an analytics-led organization.

Below are five advanced analytics and AI methods well established in sports that can be applied to organizations across industries.

1. Simulations

With binary outcomes like whether a last-second heave falling out of bounds goes in or not, how do you separate skill and luck? In order to optimize for future performance, sports teams realize that more data is better. And when real life generates only a few hundred data points, why not generate more?

The Minnesota Twins, currently battling for the top spot in the AL Central, worked with Databricks to run up to 20,000 simulations for each historical pitch they have on a player in near real time. This helps them better understand the likelihood of success for particular batters based on pitcher, pitch rotation, speed, placement, and weather conditions, which their baseball operations then uses to more accurately forecast future player performance.

Running tens or thousands of Monte Carlo Simulations has been a standard method in finance for understanding value at risk or portfolio performance. More advanced manufacturers and retailers have also used simulations to generate more accurate demand forecasts, but this was a time- and compute-intensive process due to the serialized nature of demand forecasting tools available. But cloud computing now makes it cost-effective to spin up short-term processing power, and parallelizing data tools can run millions of simulations in a fraction of a second, allowing for 100 or 1000x more levels of granularity in half the time.

2. Computer Vision and Natural Language Processing (NLP)

Computer vision and NLP automates the conversion of unstructured data like video, audio, and images, into structured data that a machine can now analyze to discover patterns. Hawk-eye Innovations, for example, provides ball tracking technology used in professional tennis, soccer, cricket, rugby and baseball used for better officiating and broadcast enhancement.

What does an in-play tennis ball have to do with steel mills, snack foods, and hedge funds? Cheetos is rolling out an AI system based on computer vision inspection system to ensure “all Cheetos to have a similar crunchy texture, a uniform shape and the same density, or airiness.” And over the last few years, hedge funds looking for alt data sources have used geospatial images to measure foot traffic to stores to forecast sales and earnings.

3. Player and Ball Path Optimization 

Given dynamic variables like opponent, positioning, and weather, where is the optimal place any individual player should go to maximize their team’s chances of success? Second Spectrum provides a deeper layer of understanding of the options and expected results of dynamic gameplay like the likelihood of a particular player to score from a particular spot based on every shot they have taken in an NBA game given the defense (e.g., man-to-man versus zone versus double team) on him at the moment. On the other side of the ball, it helps defenders understand which lanes to cut off to decrease the likelihood of an opponent scoring as the ball moves.

Delivery and logistics teams are embracing AI for path optimization using real-time conditions around weather and traffic. In warehousing and grocery delivery sectors, walking path optimization for workers can provide an outsized impact in productivity as they shift from heuristic methods to a path mapped and optimized based on the location of individual items for the specific order they are retrieving.

4. Telemetry 

Ubiquitous connectivity and sensors take the guesswork out of understanding the current condition of a vehicle, but when combined with AI, you now have the ability to make predictions and recommendations to optimize for performance. Formula 1 teams already integrate telemetry data from thousands of sensors generating millions of data points every second to optimize in-race strategy like stops, tire changes, and overtakes.

The Oil & Gas sector has embraced IoT for predictive machinery lifespan, failure, and maintenance on everything from drills to pipelines. Across many of our manufacturing customers in the heavy machinery space, we have seen a business model shift as some sell the machinery nearly at cost but turn to data analysis services and consulting as the real profit engines. As a larger trend, we are helping many customers turn data from a cost center to revenue generator. Wejo, the global leader for connected car data, has created several data products to help their customers, ranging from global data providers to city traffic planning commissions, understand traffic patterns, vehicle movements, and driver behavior.

5. Player Health, Training, and Performance Optimization

(Jacob Lund/Shutterstock)

A team’s investment in an athlete goes beyond the payroll when a star can add tens of millions or hundreds of millions to the value of a franchise. Local broadcast viewership of the Golden State Warriors fell 66% when stars Steph Curry and Klay Thompson sat out the 2019-20 season due to injury.

For sports organizations, this means an increased investment in data to prevent injuries and improve performance. Gone are the days when every professional team and star athlete had a cigarette sponsor. Lebron James alone spends $1.5M per year on his body for nutrition, training, and recovery. In 2019, the NFL partnered with AWS to help reduce player injuries. Healthcare providers and insurers are already at the forefront of leveraging data, advanced analytics, and AI to improve patient outcomes. But work sites are also seeing employers leverage sensors and AI to reduce injuries. StrongArm, for example, uses proprietary IoT wearables along with AI to derive insights from the 1.2 million data points generated per worker per day and provide a predictive measure of injury risk.

With the increased adoption of wearables, augmented reality, and edge computing, the trend will be for more in-the-moment uses of the data. That is, not only using AI before the season to optimize player selection or before the game to hone strategy, but in the game, to optimize tactics from play to play. With wearables, augmented reality, and edge computing, suddenly each individual will be able to make better decisions as the play unfolds. The advantages that accrue to data-driven cultures, whether for a sports team or a business, will widen between those who understand how to integrate data versus those who don’t.

About the author: Hector Leano is a Principal in the Industry Marketing team at Databricks, helping customers in Media, Telecommunications, Retail, and Manufacturing industries understand how to leverage data and AI to open new revenue streams while decreasing operational costs. He has decades of experience in the media industry in a variety of operational and creative roles, including at AWS, Amazon Prime Video, Sling TV/Dish, and Fox Sports.

Related Items:

It’s March, Time For Bracket Analytics

When Citizen Data Science Meets Basketball Analytics

Deep Learning Is About to Revolutionize Sports Analytics. Here’s How