Examining the efficacy of the seven metrics (NET, KenPom, Torvik, BPI, Strength of Record, KPI and WAB) used by the NCAA Tournament selection committee this year.
Bart Torvik admits it’s all a little scary.
His name is an official part of March Madness for the first time this year because the college basketball advanced statistics and rankings website he created in 2014 – barttorvik.com – is one of seven metrics featured on NCAA Tournament men’s basketball selection committee team sheets in 2025.
This started as a hobby. It’s still supposed to be.
Which NCAA men’s teams to watch on the bubble before March Madness
USAT’s Jordan Mendoza gives his teams to keep an eye on for ‘bubble watch’ as March Madness selection Sunday approaches.
Sports Pulse
Torvik, 48, is an Evanston, Illinois-based personal injury and food borne illness attorney who began down this path as a Wisconsin basketball fan and avid kenpom.com user looking for a way to isolate a Big Ten team’s performance, adjusted for efficiency only in conference games.
It’s grown into so much more.
Torvik’s website generated three million pageviews last March, and Alabama coach Nate Oats mentioned him by name at a Final Four news conference. The website almost crashed on Selection Sunday the past two years. Unlike other popular college basketball metrics, Torvik neither charges money to subscribe to his website nor does he have corporate owners. He bought the server that runs the entire operation off Amazon.
The official acceptance of the NCAA this year is an honor, Torvik said, but brings with it a new level of scrutiny and spotlight. Every team’s “T-rank” really matters now.
“I always had it in the back of my mind that I only have to be so accurate,” Torvik said. “Now, I feel more pressure.”
This sense of unease is part of an odd dynamic playing out as metrics take over the conversation in college basketball, particularly when it comes to the NCAA tournament selection process.
Phrases like NET, KenPom and Torvik have joined the vernacular of the sport over the past two decades, with fans, coaches, broadcasters and bracketologists alike placing significant stock on their rankings to differentiate between teams ahead of Selection Sunday. Meanwhile, a new metric with professional gambling roots is being introduced to the NCAA Tournament selection process this year and there’s hope among the analytics community it could bring more clarity to this complex endeavor.
But for the moment, the men who created these mathematical formulas driving how college basketball’s postseason is discussed still don’t think it’s being talked about correctly most of the time.
A flawed NCAA Tournament selection process
They were from different walks of life and different parts of the country, all together on one screen.
There was Torvik and Ken Pomeroy, the former meteorologist living in Salt Lake City who started his KenPom ratings in 2002 and watched them get so popular he made it his full-time job. There was Kevin Pauga, a Michigan State athletics administrator who created the KPI working for Spartans basketball coach Tom Izzo, as well as ESPN director of analytics Matt Morris and Alok Pattani, a Bay Area-based data science developer at Google who previously worked at ESPN.
They were brought together by the NCAA for a virtual roundtable released Jan. 28 as part of the organization’s latest effort to offer more insight into the process behind picking the 68 teams in the men’s NCAA basketball tournament.
These representatives of the seven metrics used by the NCAA tournament selection committee all agreed the NCAA improved the selection process by eliminating the Ratings Percentage Index (RPI), developing the NCAA Evaluation Tool (or NET) and embracing a variety of ratings systems, beginning with the 2018-19 season.
But they also agreed on this point: Only some of the seven metrics should actually be used to pick the 68 teams that make the NCAA tournament.
Pomeroy said his rankings shouldn’t be used. Torvik said his rankings shouldn’t be used. Nobody said the NET should be used. Not even Pattani, who helped create the NET through the NCAA’s corporate partnership with Google.
“It’s a little weird I’m on the team sheet,” Pomeroy admitted in an interview with USA TODAY Sports. “But I think everyone (on the selection committee) understands they’re not going through my rating system and picking the best teams. They understand my rating system is more predictive and you’re not picking teams based on how good they are in a predictive sense. You’re picking them based on their accomplishments.”
But fans nonetheless read and hear about most NCAA Tournament hopefuls in terms of their NET ranking around Selection Sunday, with the nuance of each ratings system often lost in the emotions of March Madness and whether a team is perceived to be ranked too high or too low.
The seven metrics on NCAA team sheets are technically divided into two categories. The NET, KenPom ratings, ESPN’s Basketball Power Index (BPI) and Torvik ratings are considered predictive rankings, or how good a team is based on its offensive and defensive efficiency, adjusted for opponent strength and location. ESPN’s strength of record, the Kevin Pauga Index (KPI) and wins above bubble (or WAB) are results-based rankings that judge how hard it was for a team to attain its resume.
Torvik and WAB are making their debut on NCAA Tournament team sheets, with particular interest being paid to the WAB because creator Seth Burn believes if selection committee members “just use that, they can simplify it quite a lot,” he told USA Today Sports, “and it will guide them in who they should select.”
Though the general principles used to formulate these metrics are made public, the exact formulas used for them are not. It’s viewed as proprietary information, even though “most of them are pretty similar,” Morris told USA TODAY Sports. “They’re using a lot of the same input data. … We’ve converged to some degree.”
Everybody in the NCAA-produced roundtable said results-based metrics are what should be used to choose teams for the NCAA tournament. Whether that’s the opinion of the selection committee remains nebulous.
The 12 members of this year’s men’s NCAA tournament selection committee, which includes nine sitting athletic directors and three conference commissioners, either declined comment or did not respond when USA TODAY reached out asking how metrics would be used during the selection process.
“The committee has more information than ever before at its disposal,” NCAA director of media coordination/statistics David Worlock said in an interview. “As a staff, we’re trying to educate them on what all these numbers mean, how they’ve been used in the past, what’s truly important when evaluating a team.”
Worlock, who has worked with the selection committee since 2006, added the committee “is using the data more effectively, in my opinion, in recent years.”
The NCAA emphasizes the NET is just its primary sorting tool, not an end-all, be-all ranking. The selection process still revolves around the human element of 12 committee members casting a vote on each at-large team that makes the field and “it’s who did you play, where did you play, how did you do?” North Carolina athletic director Bubba Cunningham said earlier this week during a teleconference in his role as this year’s men’s NCAA tournament selection committee chairman.
“And then ultimately, really good advice that I received from some of the other committee members over the years is when you get down toward the end of those last couple of teams, kind of take a step back from the metrics. Say ‘who is the better team?’ ”
Worlock acknowledged, however, the NET gets more attention from the public than other metrics because “it’s affiliated with the NCAA.”
“Some of the fault lies with the people in charge in that why are these ratings on the team sheets if they’re not being used, and I think the fact is they are used, especially the NET,” Torvik said. “People say the NET is a sorting tool, but it’s not completely true. … If you’re trying to do bracketology, you can’t just ignore a team’s NET. It does matter.”
How the NET changed NCAA bracketology
Another meeting of the minds happened in Indianapolis back in 2017. “A bit of a summit,” Worlock called it, with Pomeroy and Pauga, sports reporters, bracketologists, selection committee members and NCAA officials discussing how to modernize the metrics used to help the committee during the NCAA tournament selection process.
KenPom’s ratings based on adjusted efficiency and points per possession had become increasingly popular with coaches and fans after former Butler coach Brad Stevens said he scouted upcoming opponents in the NCAA tournament using Pomeroy’s website during the program’s Final Four runs in 2010 and 2011. The NCAA was still using the outdated RPI, which relied on winning percentage and opponents’ winning percentage to calculate strength of schedule.
The NET was born from there as an attempt to fuse predictive and results-based elements together in a single metric. It was met with skepticism. Statistician Nate Silver, then working for ESPN’s 538 website, called the NET “the worst rankings I’ve ever seen in any sport, ever” after its debut in November 2018.
But the NET quickly became the centerpiece for all bracketology discussions in recent years. Whereas teams were once tracked by wins and losses over the RPI top 50 or top 100 teams previously, the NET led to the four-quadrant system based on an opponent’s NET rating to differentiate the quality of wins. A team’s quad one and two wins, and quad three and four losses, are dissected each March. It’s not possible without the NET.
The view of how that gets digested by the selection committee, and the potential flaws, varies even among the sport’s most entrenched figures.
“The people in this room go over all this stuff, they get confused and they rely and fall back on, ‘Well, they had six quad one wins,’” said former Syracuse coach Jim Boeheim, who’s currently working as an analyst for the ACC Network. “Well, you can’t just discount if you’re good in November and bad in March. You’re not gonna be good in the tournament.”
The NCAA tweaked the NET formula ahead of the 2020-21 season, with the most notable change being the use of statistics adjusted for efficiency instead of raw statistics from a given game. It also took out winning percentage and adjusted winning percentage, which were from the RPI. Pomeroy, Torvik, Pauga and ESPN have all also tweaked their formulas from the original form over the years.
The NET does not include any preseason data or scoring margin (other predictive models do) and weighs every game the same, regardless of date. Though the NET includes a team value index component that’s results-driven, data has shown the metric tracks more closely to other predictive models, according to Worlock.
Clemson coach Brad Brownell said last year he thought Big 12 teams were able to “manipulate” the NET ratings by putting together easy nonconference schedules to boost their efficiency numbers. But the NCAA believes the NET’s relative alignment with KenPom, Torvik and BPI “indicates we have a metric that’s serving its purpose,” Worlock said.
“I think people figured out how to game the RPI so that became a phrase that people would use, that people were gaming it,” Worlock said. “You can game any predictive metric by winning a bunch of games by a lot of points.”
Others in the statistics community are still coming to grips with the NET’s creation, implementation and imperfections.
“I’m a little conflicted because I do think it’s great that they were looking to move beyond the RPI. It was time to do that,” Pomeroy said. “But at the same time, we weren’t really included in that process after that meeting. They went to Google and came up with a formula that’s almost entirely based on offensive and defensive efficiency, so you ultimately end up with a formula that’s similar to mine.”
Added Morris, who referred to the NET as “inferior” because it doesn’t include scoring margin: “If you wanted to use one of the metrics to make money in Vegas, you would not use the NET rating. Just to be blunt. It’s way better than RPI was, but it still misses out on some storylines.”
ESPN bracketologist Joe Lunardi told USA TODAY Sports the selection process is more “metric-dependent” than when he started projecting NCAA Tournament brackets in 1995. But he sees it as a reflection of the selection committee’s composition – “There used to be more basketball people on the committee way back,” is how Boeheim put it – and how much more is known about the process today.
“Are there outliers with the NET? Yes,” Lunardi said. “But fewer of them, and particularly if you cross-check them with the other popular metrics of the day, and I look at all of them. I’ve always looked at all of them. I’ve always aggregated them even before they were on the team sheets. I might weigh them differently year-to-year based on what way I think the wind is blowing.”
“If there’s a lasting legacy of bracketology, or me, or some of the early practitioners,” Lunardi added, “it is bringing to the public the process before the NCAA was either willing or able to do so.”
The proliferation of available data has helped transform bracketology from a seasonal niche to a year-round cottage industry.
The Bracket Matrix, a website that tracks bracketology experts, began in 2006 with fewer than 25 websites listed as putting together a mock bracket ahead of Selection Sunday. It’s tracking more than 90 bracketologists around the country this year and its rankings feature 179 websites or experts that have released bracketology predictions over the past three seasons.
“Part of the reality here, it’s not that there’s all these people who are brilliant bracketologists,” said Pauga, who also has a separate business built around a platform and algorithm called Faktor that helps conferences put together schedules. “I’m not trying to undermine them in any way. It’s just the process is more scripted. It’s more predictable. You can just, because of the data and the public nature of the data, you can have a pretty good idea of where you’re at. The surprise on Selection Sunday is more about who you’re playing than who’s in the tournament.”
Did a gambler discover new ‘gold standard’ metric?
They weren’t sure Seth Burn was his real name. Only that it was in his social media handle. But Pomeroy, Torvik and Worlock all brought him up independently while discussing the other new metric approved by the NCAA Tournament selection committee for the first time this year. Some believe it could help solve this convoluted conversation that has accompanied the growing acceptance of statistical formulas.
WAB, or wins above bubble, “shows how many more, or fewer, wins a team has against its schedule versus what a bubble team would expect to have against the same schedule,” the NCAA wrote in November. “The WAB metric uses NET as the basis for opponent strength, with the reference ‘bubble team’ being defined as a team ranked 45th in NET, based on a study of recent seasons.”
What the NCAA doesn’t mention is that the metric appears to have originated from a professional gambler in Bronxville, New York.
Burn, 47, stopped working as an accountant 10 years ago because he was so successful at betting on the NFL and college basketball. He’s a self-described math nerd who once used Pomeroy’s website to “crush over/unders” and realized he was “good at analyzing data and generating better projections than are publicly available,” Burn told USA TODAY Sports.
Burn first mentioned WAB and the basis of its formula in a post on his personal blog on Feb. 1, 2015. A decade later, Pomeroy views WAB as a metric that’s “getting closer to the gold standard of selecting teams” and removing human bias from the equation.
“It levels the playing field a little bit” for teams with fewer quad one opportunities, explained Worlock, noting certain members of previous selection committees have used WAB in the past to help determine the NCAA Tournament field.
If it were up to Pomeroy, there would be just one metric to determine the teams that make the NCAA Tournament for men’s basketball each year. Burn thinks he already has one. He’s just not sure what this NCAA tournament selection committee will do.
So what might happen next is a little scary.
“We’re going to know if they pick a team in the tournament ahead of a team from their very own metric that has a better resume,” Burn said. “If they do that, that’s going to be the tricky thing for them to explain. If they don’t do it, then we’ve achieved victory. Then WAB has won. That’s what I’m hoping for.”
Follow Mark Giannotto on social media @mgiannotto and email him at mgiannotto@gannett.com.
