One of the most accomplished analytics groups in sports isn’t actually in sports. At least not in the usual way.
Just three years after co-founder Doug Fearing left the Los Angeles Dodgers’ research and development department, one of the largest any club has assembled, the group of R&D experts Fearing has amassed at Zelus Analytics already rivals it in size. If Zelus doesn’t have the largest headcount yet, at 26, it may soon.
Advertisement
“I should preface that I have heard rumors about soccer teams in China with 100 people,” said Luke Bornn, a co-founder of Zelus. “But I have yet to have such rumors validated.”
Based in Austin, Texas, and backed by RedBird Capital, the company is already different from typical R&D groups in at least one way: its independence. Zelus is not housed by any one team, instead relying on a cost-sharing model in which a limited number of teams — in baseball, soccer, and soon basketball — can use its platform.
More than size, then, what will determine Zelus’ success is whether it can rival the best in quality.
“We need to be able to go to teams and convince them that we can compete with the top teams in the analytics space and their internal groups,” said Fearing, who first worked with the Tampa Bay Rays’ R&D department before heading to Los Angeles.
By this point, almost every major professional sports team has entered the world of analytics. Somewhere inside your favorite team’s front office is a group of engineers, researchers and programmers who wouldn’t be out of place at Google or Apple. The size of such staffs varies, depending on the franchise and the sport. Some might be bare-bones, while others might not be formally labeled “R&D.” But all of them are running in the same race, the one that took on a new urgency when “Moneyball” was published two decades ago: Finding new ways to win through innovation.
Cataloging all the ways in which analytics and its interlocking subjects have changed even baseball alone — for both good and bad — is its own conversation. What is certain is that inside all sports, the search for new information has grown in importance, cost and specialization, and that trend is only accelerating.
Enter Zelus, which hires expensive, experienced and lavishly educated specialists. While a specific project might require cutting-edge expertise in computer vision or spatio-temporal analysis, those needs are often too limited for teams to justify hiring the relevant expert full-time. The Zelus model brings these competencies under one roof, offering access to a range of top-end talent for far less expense than hiring the same experts individually.
Advertisement
The company then sells the group’s knowledge and mathematical models to a capped number of clubs in Major League Baseball and professional soccer: six in MLB at maximum, one per division; and one per league in soccer.
“We very much want to be focused on creating a competitive advantage for the teams that we’re working with, and to have all of the incentives line up around that. And so 30 wouldn’t work,” Fearing said. “We’re trying to do all of the same things, all of the same models, that are being developed within a front office. Where I think we’re able to create an edge is doing that in a way that is a little more rigorous and cohesive in terms of all of these models tying together.”
Despite its headcount, Zelus appears to have been extremely selective in their hires. A number of professionals with demonstrated expertise from either current or past teamside work have stumbled at the technical stage of Zelus’s interview process, according to conversations with The Athletic.
Almost a decade ago, Bornn and Fearing overlapped for about a year while teaching at Harvard, although neither imagined starting a company then. As Fearing forayed into baseball, Bornn wound up in jobs leading analytics for the Italian soccer club A.S. Roma and the Sacramento Kings. The rest of Zelus’ roster includes former executives from different sports, and leaders in fields some fans may know little about, such as machine learning, causal inference and spatial-temporal data analysis — the last piece one that is particularly important as Zelus seeks to expand into sports such as soccer and basketball, where player- and ball-tracking data are the most fertile sources of new insight.
The ability to transform what is essentially a flipbook of dots moving around a field into sport-specific words and concepts requires a level of expertise highly in demand in the broader tech space. There aren’t many who can do it, and there is plenty of competition for their services.
Advertisement
Some with this skillset, like Bornn, never intended to work in sports.
“I kind of stumbled into it haphazardly,” Bornn said. “My expertise was in modeling, spatial-temporal dynamics. Things like how fish school together, how herds of bison interact or how climate systems move. So when I got into sports, it was not because I had some great love for basketball or soccer, but because I had the technical skills to do with the new player tracking data that had come out. … When I saw that data, I saw, hey, this is like really incredible, scientific data.
“Maybe the simplest thing to think about is that traditional people in sports were dealing with box-score and play-by-play data. Which is on the order of probably hundreds of data points per game.”
Meanwhile, a single game of player tracking data in the NBA produces around a million data points, multiplied by 1,230 games per season, plus the playoffs, with seven full seasons of data prior to this one. “The types of methods or techniques you would use on that data is just fundamentally different,” Bornn said. “And a lot of people who had traditionally worked in sports analytics hadn’t yet built up those skill sets.”
Analytics, in short, have expanded far beyond Jonah Hill-as-not-Paul DePodesta staring at spreadsheets in a broom closet. Database management, exercise science, data visualization and application design are all part of the umbrella now.
For Zelus, Bornn said the ideal approach is getting the relevant experts “all in a room together and really getting them to solve the big problem of, ‘How do we create more wins?’”
“A team that’s really focused on let’s say, reducing injuries, is going to make a lot of really adverse decisions on winning,” Bornn said. “It might reduce the chance of that player being injured, but the net effect on wins might actually be negative. But if you put all these skill sets together and … get these people working together, then you can actually get to a point where you’re making decisions that are collectively optimizing toward the same thing, which is ultimately winning championships.”
Zelus does not see itself as a full replacement to a team’s in-house R&D staff, but a supplement. The reason, Fearing said, is that it’s always helpful for teams to have analysts in-house who can communicate and translate their information, be it to the players or others on staff. The greatest models and fanciest numbers in the world aren’t a big help without implementation.
Advertisement
Not every team that signs up with Zelus in a given sport is going to see the exact same thing. Zelus provides the engine, the model that takes data and then interprets it. But every team has different sets of data.
“The best way to think of it is we’re creating a Zelus database, or Zelus dataset, that they can incorporate into all of their existing tools,” Fearing said.
One of Fearing’s hires at the Dodgers, Jimmy Buffi, has struck on his own in an adjacent space. Buffi founded Reboot Motion, dedicated to the motion-capture side of sports, kinematics, which can be used to help prevent injuries and improve performance. Still friends with Fearing, Buffi has gotten a look at Zelus’ work.
“Something I really appreciated about Doug’s focus in what he was and is providing is the academic rigor,” Buffi said. “I feel like there are far too many — charlatans might be too strong of a word, but you know, a lot of people (who say), ‘Here’s our proprietary platform that pops out one number that tells you everything you need to know.’ And I’m always like, ‘I don’t know what that is, you’re not willing to tell me how you got it.’”
Citing confidentiality, Zelus declined to identify any of its partner teams save for a French soccer club which RedBird has a majority stake in, Toulouse F.C.
“Whenever we invest, we like it to lift all boats,” said Alec Scheiner, a partner at RedBird, which leaves Zelus’ day-to-day operation to Fearing and Bornn. “So the idea here is that we can invest in multiple properties and those properties can have a positive impact on each other. … Zelus can certainly help us make great decisions at Toulouse, for example, a team where we control decision-making. And then over time as we make more and more investments in teams, we have Zelus as a potential source for information and knowledge (for those teams). But we also believe in the Zelus business model.”
Zelus had all six MLB slots filled in 2020, although only four slots are filled at the moment, and is known to be looking for a new American League East partner. The Rays were a partner team previously, multiple rival executives said, but no longer.
Advertisement
One general manager whose team is not partnered with Zelus said the Oakland A’s are working with the company. The GM noted they believed the A’s, after birthing the Moneyball era, had actually fallen behind as innovators. (A business relationship between Oakland and Zelus also would be logical on another front. Bornn is teamed up with Oakland’s Billy Beane in a separate endeavor tied to RedBird, a special purpose acquisition company, or SPAC, called RedBall. Beane is not involved in the management of Zelus, however.)
Another GM with a club that does not work with Zelus said they understood the White Sox to be Zelus’ American League Central partner, and that the San Francisco Giants were its National League West partner. They pointed to the Giants, who still have an up-and-coming R&D department, as interested primarily in help with modeling.
Multiple rival executives said that the Atlanta Braves were also a partner of Zelus’.
The drop from six MLB teams to, for now, four, doesn’t have Fearing concerned, he said. Bornn noted how hard MLB teams were hit by COVID-19, which became a crisis in the U.S. just as the 2020 season was to begin.
“I actually feel really good, particularly given the broader economic context, that we got four of our six teams to renew at higher rates than they were at before,” Fearing said.
Now in its second full season working with MLB clubs, Zelus is eyeing expansion into other sports. The company didn’t plan to launch in basketball until 2022, but Bornn said interest grew when MLB owners started recommending Zelus to counterparts in the NBA, and the kickoff for an NBA platform is targeted for Aug. 1. The company has already reached out to every NBA team, and Bornn said the demand has exceeded available slots.
Analytic prowess and needs vary widely across sports. With their average analytics staffs numbering four to five, basketball R&D departments are far smaller than those of powerhouse baseball teams. In equal measure, the difference is a reflection of the youth of the analytics movement in basketball, as well as the greater complexity of managing larger roster sizes in baseball. MLB teams have a stable of minor league players as well, plus a draft that includes many more rounds than the NBA’s. But to an extent, Zelus’ specialized techniques and methods are applicable across sports.
Advertisement
How much does Zelus cost? On confidentiality grounds, Fearing declined to say. The NBA product, sources said, is being offered at an annual rate in the low six figures. One baseball GM put the cost at $500,000 or $600,000.
“What I can say is that our model early on with all these sports is to operate at cost,” Fearing said. “We’re trying to take every dollar that we get from our partner teams and put that into building out a competitive advantage in return through the sports intelligence platform.”
Unsurprisingly, a variety of opinions can be found about a start-up that’s doing business differently.
“It’s really good for teams that are either way behind or kind of in the middle analytically and need to play some catch-up,” one MLB general manager said, expressing fondness for Fearing, while noting teams would prefer to own their mathematical models.
In the NBA analytics world, the Zelus pitch has been received with a mixture of interest and skepticism, per conversations The Athletic had with team officials. Bornn, with prior NBA experience, is as respected a practitioner as there is in the industry, and his involvement gives the company legitimacy as it enters a new space.
Sources noted the annual cost is comparable to, and perhaps exceeds, the premium level of access to Second Spectrum, to which the overwhelming majority of teams subscribe (on top of the raw tracking data which Second Spectrum provides to each team via a partnership with the league).
Some NBA officials questioned the apparent focus on player valuation metrics; minor improvements on teams’ internal models would not be nearly as valuable as gleaning the sorts of insight from tracking data that require the level of technical skill at Zelus’ disposal, according to those sources.
What does an ideal partner team look like? Teams with larger analytics staffs with dedicated long-term research arms might perceive the offering as superfluous, while for those on the low-end of current investment, there could be insufficient in-house expertise to receive full value.
Advertisement
In theory, Bornn said, the team that would benefit the most from Zelus’ offering is the team with the least invested in analytics.
“Because we can very quickly get them on par with the smartest teams in the league, and within a few years actually put them above the smartest teams (with) very minimal investment because of the cost-sharing model, you can make massive strides with those teams,” Bornn said. “Teams that are invested the most heavily see us as a way to get additional capacity and push their methods to sort of new highs. … And then we see teams, mid-tier teams, that don’t have the time to think about more in-depth modeling tasks, and hey, this is a way for us to really up our game and get the top level of insights.”
In both sports, executives wondered if the transparency Zelus is offering would sufficiently help in situations where a decision-maker is torn, citing the importance of “having somebody who was not just available, but in the office next door to explain it,” as a baseball official experienced in R&D put it.
Interestingly, Zelus plans to be open about its models with its partner teams, on the belief it would be very difficult for a team to effectively catch up on its own in a way that would be damaging.
“We share with our partner teams white papers that describe the full statistical detail of (our) models, and the reason for doing that again is to empower that internal group,” Fearing said. “We don’t want it to be a black box. That’s been a very explicit decision. There’s some trade-off there. I think there are a lot of companies that try to keep everything black box because they feel like that is their secret sauce, and don’t want to give anything away and maintaining that proprietary information set is important to their success. Whereas we’re really trying to be successful based on what we continue to develop. The cutting edge continues to move.”
R&D groups, in sports and otherwise, can be very hush-hush because businesses want to protect their advancements and intellectual property. Fearing said that Zelus keeps the data of every team it works with separate and proprietary.
One baseball GM wondered about Zelus’ staying power, but overall, reactions were positive.
Advertisement
“It might take a season or two, but I’m sure they will figure it out,” said one NBA team’s head of analytics.
“In a lot of ways, it’s a really inexpensive way to validate the stuff that the teams are learning themselves,” Scheiner said. “And if you think what’s at stake for these teams when they make player-related decisions, so much money at stake, that spending a little extra on people like Doug and Luke to validate or to challenge some of the assumptions you’re making is always going to be worth it.”
From one other angle, Zelus’ arrival appears notable. Its structure provides a way for people to work in sports but without having to commit to the lifestyle of working directly for a team. One of the reasons Fearing started the company in Austin, where there is no MLB, MLB, NFL or NBA team, was because his son was starting high school there.
Zelus also allows its people to do a different type of work. Fearing’s approach is proudly academic, but the day-to-day demands of a front office and a baseball schedule don’t always line up with that M.O.
“There isn’t a great technical career path within a front office,” Fearing said. “There are some sets of people who I think just fit better at Zelus. Because they like that research context, they like the idea that they can work on multiple sports, they like having a group around them who also have that type of mindset.”
More recruiting may follow. As it expands into basketball, the company could eventually expand beyond pro sports, into the amateur space, telecasts and gaming.
“One lever to pull is to move beyond baseball, and we’re doing that and I think that’s very comfortable,” Scheiner said. “The next lever is to move beyond the team model and look at things like broadcasts or other things where people can use this analytical approach and our sophistication. And the last one would be even if you get good enough at digesting big data, is move beyond sports.”
Advertisement
Scheiner, formerly the president of the Cleveland Browns, said he expects Zelus to expand to the NFL as well.
“If we get to a point in three to five years that we have the best predictive modeling capabilities for sports, period,” Fearing said, “there’s going to be a lot of doors that open.”
(Top photo: Brad Mangin / MLB via Getty Images)