Carnegie Mellon Revolutionizes Sports Analytics: Pioneering Data-Driven Insights for Competitive Edge

Andrew Corselli

(Image: StockSnap via Pixabay)

Today’s sports analysts have access to more — and better — data than ever before. Carnegie Mellon University experts are turning that data into insight, using statistics and data science to help professional teams gain a competitive edge.

“Every tenth of a second, the NFL’s Next Gen data chips provide information for where every single player is positioned on the field — the direction they’re moving, the speed they’re moving,” said Ron Yurko, Assistant Teaching Professor in CMU’s Department of Statistics & Data Science and Director of the Carnegie Mellon Sports Analytics Center.

Tracking players on the field extends beyond the NFL.

Ron Yurko (Image: CMU)

“The MLB has information about every single swing in Major League Baseball,” said Yurko, who is also an academic partner with the NFL. “In baseball and basketball, they have what’s called ‘pose skeletal data,’ where we know at every fraction of a second, where is the elbow, the shoulder, the kneecap, and in three-dimensional space.”

Of course, the question everyone wants to answer — from team owners, managers, and coaches to analysts, bettors, and fans — is what to do with all of that data. Read on for an exclusive Tech Briefs interview, edited for length and clarity, with Yurko, who knows exactly what to do with the data.

Tech Briefs: What kind of technology do you use to collect this data? And can you please explain in simple terms how it all works?

Yurko: We don't personally collect the data at Carnegie Mellon; that data comes from the NFL.

The NFL has chips in the shoulder pads of every player on the field. Those pick up where they are on the field every tenth of a second; how fast they’re moving; the direction in which they're moving and changing; how they’re accelerating over the course of the play. Then, as researchers, we'll get access to that data — whether it's through collaborations with the NFL or releases of the data that they provide — and we'll work on it. What new insights can we gather from this? And our students, they'll work on these projects, they'll enter the NFL's annual Big Data Bowl competition, and then this leads to job opportunities.

And at the college level, what happens is they don't have the chips in the shoulder pads of college football players, but you have an amazing amount of video. And using computer vision AI technology, we can convert that video into very similar type of data that comes from the chips in the NFL. What NFL teams are working with now is the same type of rich tracking data across all of college football; they're using that to figure out who are the players they should draft.

Tech Briefs: Can you talk about some of the chips that they’re using?

Yurko: At the NFL, these are RFID chips — infrared signals that are being picked up within NFL stadiums. It's local positioning systems that really give extremely rich data that the NFL has. Beyond that, now the NFL has joined the likes of Major League Baseball and the NBA, where they have full camera systems around the stadium. And the NFL in this past year, they use this camera system, I think it’s Hawkeye, to pick up where the ball is to mark the first down — does the ball cross the first down line — rather than dealing with the old-fashioned chains. Now that can be automated because of how they could pick up the football.

From that same technology, the NFL now has access to where are literally the hands, the elbows, the knees of athletes — a full skeletal representation of football players on the field at fractions of a second. That technology now exists and is being used in the NFL. It's been used by Major League Baseball teams and NBA teams for a number of years now, but we're seeing it see now across other sports.

Tech Briefs: Which league, would you say, has the most advanced tech?

Yurko: I would say, at this point, Major League Baseball. It’s probably been the most advanced from the use of technology and statistics — “Moneyball” being the famous story from back in the early 2000s about the Oakland A's usage of data.

So, Major League Baseball teams were definitely the earliest adopters of thinking about how we integrate technology and use statistical modeling, machine learning, to understand how to evaluate players. You see it not just by teams but also players themselves.

Players will go to these pitching labs — they're using high-speed cameras, understanding what happens when they modify the grip of the baseball; how that changes the forces and the flight path and the movement; designing new pitches. And they're literally using all of this technology and data themselves to become better players. So, baseball is ahead relative to other sports.

In the NFL when this tracking data from the chips started in 2017, that led to a little bit of this arms race, where now NFL teams were starting to hire data scientists, machine learning engineers, to really figure out how do you work with all that information.

Tech Briefs: You're quoted in that article I read as saying, “Literally every tenth of a second, the NFL’s Next Gen Data chips provide information on every player’s position, speed, and direction. The question now is what to do with all that data.” Well, my question is: What's the answer? What do you do with all that data? What are your next steps? Where do you go from here?

Yurko: That’s a good question. What we've been working on as researchers, and what we know is being used by NFL teams, is how do we characterize the movement that we're observing and understand what is high-value type of movement, high-value positioning — for instance, which defenders are playing the best coverage against receivers as they're running routes. These were a type of statistic that we didn't have access to before in traditional box scores. If you had a defender covering a receiver, and if they were this amazing defender that a quarterback never attempted to throw their way, they would never get an interception, they would never get a tackle, they would never have any counting stats.

Transcript

00:00:02 If you consider football being like a dinner, scouting, coaching, that's the steak. Sports analytics to me it's the mashed potatoes. It's the gravy. It's something that can make a good meal a great meal. They've been scouting at the combine for the last 40 years. How far can you jump? How tall are you? What's the new thing we can figure out? When do we expect this person to be drafted? The analytics are behind virtually every aspect of the game. CMSAC is the Carnegie Mellon Sports Analytics Center,

00:00:28 which is our hub for research and education and different events that we have at Carnegie Mellon University. Our students have become really like the standard of what type of individuals professional sports teams and leagues are looking for. We now have access to fractions of a second where every player is at on the football field due to chips that are in their shoulder pads. So we literally know every 10th of a second how fast they're moving, the direction they're moving, what's the orientation of their shoulders. So what we have focused on as a group is building up statistical methodology,

00:01:02 machine learning methods, AI tools that can lead to actionable insights. Given this data set, can we be creative enough with our analysis so that we can identify useful insights to create competitive edges for a team? Within a play, we can literally predict where they're going to end up on the field, how many yards are they're going to gain. What's the probability that they get a touchdown. Right. That all requires pretty sophisticated methodology. Sports means so much to so many people. Data gives just another way to connect with the sport.

00:01:35 Being able to take unstructured data and dive into it and tell a story is super rewarding. It's amazing getting to see CMU students turn their passions into careers that, you know, help shape the future of professional football. If you consider football being like a dinner, scouting, coaching, that's the steak. Sports analytics to me it's the mashed potatoes. It's the gravy. It's something that can make a good meal a great meal. They've been scouting at the combine for the last 40 years. How far can you jump?

01:00:19 How tall are you? What's the new thing we can figure out? When do we expect this person to be drafted? The analytics are behind virtually every aspect of the game. CMSAC is the Carnegie Mellon Sports Analytics Center, which is our hub for research and education and different events that we have at Carnegie Mellon University. Our students have become really like the standard of what type of individuals professional sports teams and leagues are looking for. We now have access to fractions of a second where every player is

01:00:51 at on the football field due to chips that are in their shoulder pads. So we literally know every 10th of a second how fast they're moving, the direction they're moving, what's the orientation of their shoulders. So what we have focused on as a group is building up statistical methodology, machine learning methods, AI tools that can lead to actionable insights. Given this data set, can we be creative enough with our analysis so that we can identify useful insights to create competitive edges for a team? Within a play, we can literally predict where they're going to end up on the field, how many yards and they're going to gain.

01:01:25 What's the probability that they get a touchdown. Right. That all requires pretty sophisticated methodology. Sports means so much to so many people. Data gives just another way to connect with the sport. Being able to take unstructured data and dive into it and tell a story is super rewarding. It's amazing getting to see CMU students turn their passions into careers that, you know, help shape the future of professional football.