During the Big 12 tournament semifinals, Kansas guard Remy Martin took a mid-range fadeaway from about 17 feet. How the coaches on each bench felt about that shot when it left his hand varied, based on their reliance on analytics.
If they subscribed to the website hoopmath.com, for instance, they would know that a 2-point jumper from Martin went in at a 47.1 percent clip. If they were on Synergy, which tracks every shot and play type in each game, they’d know that Martin made 49.2 percent of his jumpers off the dribble and shot 50 percent on jump shots from 17 feet to the 3-point line. Since the jumper occurred in a Big 12 game, the Jayhawks coaches could assign a value to it using ShotTracker. That company, which is located in Merriam, Kan., has a chip in the ball and inside the jersey of every player. At ShotTracker headquarters, the game plays out with the players showing up as dots on a screen. Kansas coach Bill Self helped come up with what statistics — like paint touches and ball reversals — that the company tracks.
Advertisement
The KU coaches also had at their fingertips analysis from Jam Basketball Intelligence, a third-party analytics company, which provides opponent and self scouts. JBI’s report on Martin said he made 55 percent on those type of shots last season and was 13 percent better on contested looks verse uncontested. And if the coaches subscribed to ShotQuality, which gives an expected value of every shot taken in a basketball game, they’d know that shot was worth about 0.58 points.
🤩Season Highlights🤩
FIRST YEAR PARTNER @KUHoops #JamFam🍇🏀
🔵National Champions
🔴Big XII regular season & tournament champions
🔵Number one ranked eFG% defense in the Big XII
🔴Most wins in 10 years
🔵Held teams under scoring average in 9 of their last 10 games#RockChalk pic.twitter.com/5X3OIvwEfP— Jam Basketball Intelligence (@JBItweets) April 18, 2022
Simon Gerszberg, the creator of ShotQuality, used that Martin field-goal attempt as part of his presentation for prospective clients this offseason. On the video, you see a bubble over Martin that shows the expected value of the shot, and at the top of the key you see teammate Christian Braun wide open with a 35 percent bubble over his head. The video freezes, and lines point to both Martin and Braun, showing that while Martin’s shot was worth 0.58 points, had he passed to Braun, that shot would have generated 1.05 points on average.
And had Kansas coaches later tried to tell Martin he should have actually passed, they would have had the power of film and math behind their message.
This is college basketball in 2022.
Over the last decade there has been an analytics boom in basketball. In the NBA, teams have built their own analytic staff and keep their data in-house. On the college side, few programs have the budgets to copy the NBA model, which has created a market for outsiders to cash in.
Advertisement
The Godfather of college basketball advanced stats is Ken Pomeroy, whose site kenpom.com is gospel for some and recognized by most who work in the game. Pomeroy’s site became so popular that he was able to quit his job as a weatherman in 2012 and focus entirely on hoops. But he’s no longer alone in the space. Not only is there a demand for data, but there’s also a market for explaining it in actionable terms.
Colton Houston, who co-founded analytics firm HDIntelligence, uses the example of “Moneyball,“ the best-selling book that later became a movie.
“A lot of people’s takeaway from that was on-base percentage is undervalued,” Houston said. “But that wasn’t the takeaway. That was the example in the book, right? But of course, once the book is written, and everyone knows that, that’s no longer a market inefficiency anymore for those teams. So the takeaway from the book is, what advantage can data analytics help you unearth? What’s undervalued? Where’s the opportunity to have a potential competitive advantage? Data analytics is a continual search for those advantages.”
If this movement has a poster child on the coaching side, it’s Todd Golden, the 37-year-old who was hired last spring at Florida after three seasons as the head coach at San Francisco. Golden has invested in numbers on his own staff. Jonathan Safir, who worked as an assistant at San Francisco, followed Golden to Florida and has the title Director of Basketball Strategy and Analytics.
Golden and Safir are disciples of Washington State coach Kyle Smith who, while at Columbia, came up with a system where he tracked just about everything on the floor, including hustle stats, and assigned values to everything they measured. To this day, both he and Golden use that scoring system to decide playing time.
During the 2019 Final Four in Minneapolis, Golden and Safir met with Pomeroy and asked the data guru this question: Was there anything no one was doing in college basketball that the numbers suggested they should be? Pomeroy told Golden that teams should foul at the end of the first half to gain an extra possession.
Advertisement
Nine months later in a game against Pacific, San Francisco put the idea into practice. After Jamaree Bouyea made a free throw to put the Dons ahead by 12 with 11.4 seconds left in the first half, Golden instructed his team to foul immediately. Pacific’s Pierre Crockrell went to the line to shoot a one-and-one. He made the first and missed the second. San Francisco got the rebound, Bouyea attacked the paint and kicked out for a buzzer-beating 3 by Remu Raitanen. So instead of trying to simply get a stop, which would have meant Pacific ended the half by scoring zero to three points, San Francisco got a 3-1 advantage by fouling and pushed its halftime lead to 14.
Two weeks later, Golden put the theory to test at the end of a game. San Francisco held a two-point lead with 22.3 seconds left against BYU, which entered the game as a five-point favorite, according to Pomeroy’s projections. Safir told Golden that he thought the statistical play was to foul Yoeli Childs, a 57.5 percent free-throw shooter. Childs was also BYU’s best rebounder, so the Cougars’ chances of grabbing an offensive rebound would decrease. The Dons executed the plan, fouling Childs, who missed the front end of a one-and-one. Raitanen got the rebound, was fouled, made both free throws to create a four-point advantage. San Francisco went on to win the game by one.
When COVID-19 shut the world down, Safir asked Pomeroy to help determine the statistically-correct move in those end-of-game scenarios. They created handy charts that help dictate the proper move.
Safir needed Pomeroy because he couldn’t research the topic in a timely fashion without the ability to code. Pomeroy used 11 years of play-by-play data for his research. Pomeroy was willing to help because he knew the San Francisco coaches would value the results of the study, which is available for anyone to see on his website.
“People are just risk-averse and not willing to challenge conventional norms and the way things are always done — or the way things will continue to be done or need to continue to be done — until somebody like us comes and disrupts and continues to disrupt and hopefully has success disrupting,” Safir said. “We think our competitive edge and differential ability is deciphering information that is readily available. We think we’re really good at it and can’t code and can’t create our own data, so we’re relying on the publicly available data and breaking down sites like KenPom or (talking to) Ken himself. Whereas NBA teams have entire front offices working on it.”
The new Florida coaches believe in process over results. They’ve been teaching the ShotQuality method, for instance, with their in-house stats for years. In practice, a mid-range jumper that goes in is worth one point; if it misses, it’s minus-2. A 3 that goes in is worth 3 and minus-1 if it misses. A shot at the rim that goes in is worth two and minus-1 if it misses.
“We don’t ever really discourage or disincentivize (mid-range shots) when we run dummy offense,” Safir said. “Everything is around shooting layups or shooting 3s. So our guys are kind of mindf—ed. They’re not even thinking. They’re just playing. Why would you shoot that shot? That’s a shot we don’t practice.”
Advertisement
The Florida coaches watch every practice and game together, making sure they all agree upon the numbers that are assigned to each play and player.
It’s a tedious process, a similar exercise that turned Gerszberg into an entrepreneur.
During the 2019-20 season, then-Colgate assistant coach Dave Klatsky tasked Gerszberg with tracking every shot in practice and games and assigning an expected point total based on the quality of it the attempt. Gerszberg had shown up in Klatsky’s office a year earlier, when he had randomly been assigned as freshman roommates with Colgate guard Tucker Richardson and expressed a passion for analytics in sports. Richardson told him to go see Klatsky.
Gerszberg and Klatsky came up with a system where they’d grade the quality of a shot on a 1-to-100 scale. The value was based on the location of the shot, who was taking it, how open it was and whether it was off the catch or the dribble. For instance, a catch-and-shoot 3 from a good shooter might be a 50, which would mean it’s worth 1.5 points. Gerszberg would chart every shot, then Klatsky would go back through the tape and check his work.
“I freaking hated it,” Gerszberg said. “Because I had friends on the team and I wasn’t even watching the game anymore. I was just outputting numbers on this clipboard, like all these data points spinning out of my head. It was so nauseating.”
That Christmas break Gerszberg decided he’d see if he could automate the process, using multiple play-by-play data sets to do the work for him. What his algorithm spit out was nearly identical to what he was tracking by hand. He showed the Colgate coaches his work. “They were totally in shock,” he said.
The rest of that season, Gerszberg continued tracking every shot by hand and cross-checking his automated data. It was always within one point. After the season he learned to program what he was doing on his spreadsheet for every team in college basketball and in the NBA. A business was born.
Advertisement
During the 2020-21 season, 25 teams signed up for ShotQuality in what was basically a proof-of-concept season. Last season, 60 teams paid for the service, along with 500 bettors. ShotQuality spits out a projected score of every game; Gerszberg’s model correctly predicted the over/under in 54 percent of games last season with 15 pushes, and it correctly picked the closing line-adjusted winner 51.4 percent of the time with 30 pushes.
Gerszberg bet on himself by essentially quitting school to run his business. He didn’t enroll in any classes during the first semester of what would have been his senior year and then took only two classes in the spring. He’s now a college dropout, a decision his parents weren’t too pleased with initially. Then this summer Gerszberg received over $3 million in ShotQuality’s first funding round.
“Which is part of the reason my parents are OK with me taking a little bit of a break from school,” he said.
This season he’ll offer a service that will use computer vision to track the players on the floor and extract the same data he’d been tracking from the screen, precisely telling the distance of every shot and how close a defender is to the shooter. Gerszberg hired Neil Johnson from the Washington Wizards to add the computer vision tool to his service. He saw Johnson present at the MIT Sloan Sports Analytics Conference in 2020, and Johnson convinced him to focus on college and not the NBA. The NBA, Johnson told him, has Second Spectrum, which uses cameras around the arena to track the movement of every player. ShotTracker provides a similar service for college, but only the Big 12 and Mountain West have the technology installed in their arenas.
Johnson argued to Gerszberg that colleges didn’t have anything to measure process-based data; their hope is to allow coaches to change how they deliver information to their players.
“Using analytics to coach is really good, but I think it’s even more valuable when you have film to pair with it,” Golden said. “In my mind, that’s an easier way to teach guys shot selection than trying to do it without having that film to fall back on.”
Colton Houston was the director of operations at Alabama under Avery Johnson when he became fascinated with the analytics movement. Near the end of his Crimson Tide tenure, Alabama started using alumnus Matt Dover as a consultant for scheduling. Dover worked in politics and projecting elections, and he figured his expertise in data modeling could be used in college hoops. Houston later pitched Dover on starting their own analytics firm.
Advertisement
“It seemed inevitable to me that eventually analytics would have the same type of impact in college basketball as it does in the NBA, because it’s a powerful tool if used correctly,” Houston said.
Their company, HDIntelligence, launched in April 2019. The first season, Alabama, Cincinnati, Dayton and Incarnate Word were clients. This season, they will have 11 employees and project to work with 70-80 teams. They also recently signed a deal with the SEC to consult on scheduling for all the league’s teams.
Patrick Stacy, who runs Jam Basketball Intelligence, has a similar concept. Similar to Gerszberg, he started out as a volunteer during his time as a student at Loyola Chicago. After college he took a job with CDW as a financial analyst but continued his consulting work for Loyola Chicago. After two seasons of taking vacation to travel for his volunteer job, he decided to make it a career. He worked with seven teams last year, most notably Kansas. He signed the Jayhawks when he cold-messaged assistant Kurtis Townsend with a Baylor scouting report he had built the season before.
Both HDIntelligence and JBI send scouting reports before every game and also help with different projects and recruiting work, including tracking transfers. Dover has a system that assigns a value to every player, similar to an overall rating given to players in a video game.
“When I started and wrote the code for my website, it was really laborious and difficult,” Pomeroy said. “Now if you have data and you want to figure out what things matter in certain situations, it’s pretty easy to cut that data up and come up with some meaningful insight in a short amount of time.”
College teams could try to do this work themselves, but Houston argues it’d be more expensive. And if the person in charge of the analytics is good, you run the risk of losing them to the NBA or another team.
It’s not just the data that’s valuable either, Houston contends; it’s being able to explain the data and how to put it into practice. That’s where Houston and Stacy believe they provide real value for coaches who want to use numbers to find an edge.
Advertisement
“There’s probably still a minority of coaches who are anti-analytics,” Houston said. “Ten years ago that might have been the majority of coaches, and now I think it’s a fairly small minority. If we tried to launch this company a decade earlier, we would not have had as much interest as we do now.”
Golden believes most high-major programs will have a staffer dedicated to analytics in the next 10 years, as younger coaches like himself take over programs.
That doesn’t mean the third parties will be eliminated. BYU coach Mark Pope, for instance, hired Keegan Brown three years ago to be his director of analytics and video. In addition to analytics work Brown does in-house, one of his responsibilities is helping Pope decide what outside products he’ll use. Last season BYU contracted with ShotQuality, Synergy, HDIntelligence, Just Play Sports Solutions (a virtual playbook) and Noah (a camera system that tracks the flight of a shot).
Safir sees coaches like Golden and Pope as outliers still, skeptical most coaches are truly bought into allowing analytics to help them make decisions.
“Analytics is a catchy buzzy term,” Safir said. “But it’s also kind of like how talking about religion can be a turnoff to some people. We just try and break it down into simple terms that can be easily applied and provide actionable bits of information and data to our team.”
Arkansas coach Eric Musselman believes there’s more of an investment in analytics in the NBA because the general managers hire those people, not the coaches.
Musselman has long been bought in on the data. Arkansas has its own systems in place — for instance, Musselman still uses a rebounding-per-minute metric that his father, Bill, came up with years ago — and he’s always searching for new ways to use numbers in decision-making. He makes his staff read “Moneyball” and regularly sends them stories about how other teams use analytics. His interest lately is in football’s evolution in the space. He recently went to visit the Los Angeles Chargers to find out more about Brandon Staley’s propensity for going for it on fourth down.
Advertisement
“The reason it works so well with Coach Staley is because he’s the head coach,” Musselman said. “He believes in it, and he’s bought into it. And then there’s other people where there’s a line in the sand and there’s friction between analytics and coaching staffs.”
Musselman has also studied the Baltimore Ravens’ approach to building an offense, investing heavily on the offensive line and tight ends. His interpretation of how to apply that approach to basketball is at the free-throw line. In his eyes, having a high free-throw rate is basketball’s version of controlling the line of scrimmage. An attacking team that gets to the line generates more spacing on the floor.
It’s all about creating an extra point or possession here and there, which in the long run can be the difference between a win or two. When that can be the difference between getting into the NCAA Tournament or going a round further in the tourney, then that makes the investment worth it.
So what’s next? Houston suggests AI-powered programs could pull data from a broadcast, which is similar to what ShotQuality will be doing this season. Safir said he believes rest and recovery are “ripe for disruption.”
“I think we are getting to a really interesting stage where, with so many more products being out there and so much more data being available, that it will take some skill to use that data properly,” Pomeroy said. “We’ve come a long way from just looking at four factors and interpreting what that means. We’ve now got to the point where I think you could actually use data in the wrong way. And so I think understanding maybe the limitations of what is useful and what is not useful, that is going to be kind of how things evolve going forward.
“Simply signing up for these services and thinking that’s just going to be an advantage right off is not the way things are anymore. You’ve got to figure out specifically how you’re going to use this data to give yourself an edge and make your team better.”
(Top photo of Remy Martin: Courtesy ShotQuality)
