With the FIFA World Cup underway, a lot of the data that will be gathered will be done so by a company called STATS. We talk to its director of AI, Patrick Lucey.
Patrick Lucey is director and vice-president of artificial intelligence (AI) at STATS (Sports Team Analysis and Tracking Systems), a Chicago-headquartered sports analytics company.
It emerged this week that STATS is on track to employ 150 full-time and part-time staff by 2020 at its new offices in Limerick.
‘We want to tap into that talent and domain expertise because Limerick is a sports city and Ireland just loves sport. It’s just a great place to be’
– PATRICK LUCEY
The new offices are at Riverstone House in Henry Street where analysts and researchers will provide sports fans and professionals with on-the-button, immediate insight on gameplay from more than 45 different sports, including the FIFA World Cup, which kicked off this week.
Prior to joining STATS, Lucey was based at Disney Research for five years where he conducted research into automatic sports broadcasting using large amounts of spatiotemporal tracking data.
Previous to that, he was a postdoc researcher at the Robotics Institute at Carnegie Mellon and at the University of Pittsburgh, conducting research on automatic facial expression engineering.
Interestingly, Lucey’s mother is a native of Limerick, where STATS is locating its new operation.
What role does STATS play in sports?
If you haven’t heard of STATS before, you touch us every day.
STATS has been around for more than 35 years – if you check a score on Google, they get their sports data via us. We work with all the big media companies in the world and they get their sports data from us.
How do you gather that data?
It’s a mix between human and technology. For a large part, we have humans who watch games and this is a well-oiled machine that collects data. We collect the world’s largest amount of sports data but also the best and most reliable data.
We have two main modes of collecting it: aside from humans, the other method is computer vision.
We have been pioneers for years in the computer vision tracking space, using it since 1999. Prozone was really the frontrunner in collecting football data and we acquired it a few years ago. Its DNA is in football and soccer, and it was collecting tracking data since 1999.
The tech is 20 years old it but really came to the forefront in 2008 when we purchased a company called SportVU, and we started collecting player tracking data from a computer vision system that was used by the National Basketball Association (NBA).
We were listed in Fast Company’s Top 100 Most Innovative Companies because we had this tracking technology that is pervasive in the NBA space and is really leading the charge in the next age of sports analytics.
You can think of STATS as a Google for Sports – we collect all of this data.
How does AI come into it?
My background is in computer vision and speech recognition. I did my undergraduate thesis in face recognition and then a PhD in audio visual speech recognition. I spent a large part of my time using IBM Watson for speech recognition regardless of head position.
Using an active appearance model like you would have seen in the movie Avatar, we were able to read faces. I was at Carnegie Mellon for two years as a postdoc and we used that technology to help measure human behaviour such as pain, depression and facial paralysis. I was thrilled to be able to use machine learning and computer vision as a measuring tool.
I worked at Disney Research for five years where we tried to do an automated sports broadcast. The idea there was that we are seeing a paradigm shift in how sports is being broadcasted or consumed. Now, a lot of people are watching it on mobile devices.
However, the problem is that generating content costs a lot of money and it is mostly about being surrounded with humans and bringing trucks in, laying cables and doing that for a one-off broadcast. So, cost is prohibitive and the idea was that we could set up a smart venue and for a one-time cost, set up cameras and track players in real time, and set up an automated sports broadcast.
My job at Disney was to teach a computer sport by tracking data, and that led us to a lot of great things that we are trying to do at STATS.
What kind of things can you do with AI and sports data?
Well, there is just so much. As I said before, STATS has been around for 35-plus years, and there are various modes of data that we have.
We have Box Scores, which is a quick summary of a match in terms of numbers. Then you have Play-by-Play, which is a textual description of every play. Then you have Tracking Data, and using computer vision and wearables. Then we also have a video component that captures data in real time.
The way we like to think about sports data is as a mechanism of reconstructing the story of a match. Given simple data, you can tell a story but at a very high level.
With Play-by-Play and Tracking Data, you follow it in a very granular way.
How will the new Limerick operation feed into this data-driven world of sport?
Part of its function will be operations, collecting data and also siphoning through it. But another big aspect is doing the research and understanding this data, and eventually our aim is to have part of the AI team based in Limerick to leverage all the understanding that we have, especially given the strong DNA in rugby and football in Limerick – it makes sense that we have some aspects of it there.
Sports is a very powerful vehicle to understand abstract things. What we are particularly good at in STATS is AI, and in particular machine learning. What we are particularly after are top developers who have an understanding and a passion for sport but also a passion for technology.
We are really looking forward to people with that skillset because sports is one of those rare areas where people really like it and are passionate about it but there’s a very strong technical element to it. I like to think of sports data as one of the most challenging and interesting areas in machine learning.
Sport is just one of the most fascinating areas to try cutting-edge machine learning, and that’s what we strive to do.
With the FIFA World Cup in full swing, what role will STATS be playing?
For the World Cup, we will be collecting a lot of tracking data. This involves collecting structured data, what you can do with the Tracking Data, and how teams can use this technology to get a better understanding of performance and selecting future players.
A computer can see every game that’s played, and every game will be tracked, but the computer collecting that data intelligence is still nothing compared to a human. So, what we wanted to do was help experts to do their job better.
An analyst in the States could spend 29 hours watching video just to find a suitable player.
Big data is diminishing the quality and application of this. The real sweet spot is the granularity. The granularity allows us to ask very specific questions and get answers back.
We will have products coming out from STATS soon that will allow users to search for plays in basketball, such a three-point shot. But with Tracking Data, you can be very descriptive.
A picture tells a thousand words but in sports data, what is the right language to use?
We strongly believe that sports is very visual, so why can’t we query on that?
This paves the way for visual input as a way of searching. Instead of typing words, use an image or clip of a perfect pass or play. We call this concept short-boarding and you can think of this as YouTube for Sport.
Users will be able to scrub through and find parts of matches that they are looking for. This will be great for a coach to be able to explain a tactic to the team and be able to do this in less than a second on a tablet computer.
This is where AI really kicks in – the ability to leverage the data that we have.
We want to tap into that talent and domain expertise because Limerick is a sports city and Ireland just loves sport. It’s just a great place to be.
Want stories like this and more direct to your inbox? Sign up for Tech Trends, Silicon Republic’s weekly digest of need-to-know tech news.