“The Truth Behind March Madness Selections Exposed”
Image by Ramon Padilla, USA TODAY
Bart Torvik finds it somewhat nerve-wracking. For the first time, his name is officially involved in March Madness this year, as his college basketball advanced stats and rankings website, barttorvik.com, is one of seven metrics featured on the NCAA Tournament men’s basketball selection committee team sheets in 2025. What began as a hobby for Torvik, a 48-year-old personal injury and foodborne illness attorney based in Evanston, Illinois, has now become a significant endeavor.
Torvik’s journey began as a Wisconsin basketball fan and keen user of kenpom.com, seeking a way to assess a Big Ten team’s performance adjusted for efficiency solely in conference games. Fast forward to today, and his website attracted three million pageviews last March, with even Alabama coach Nate Oats acknowledging him by name at a Final Four press conference. The website nearly crashed on Selection Sunday in the past two years. Unlike other prominent college basketball metrics, Torvik’s website is free to access, with no corporate backing – he even bought the server running the operation from Amazon.
While Torvik appreciates the NCAA’s recognition this year, he acknowledges that it comes with heightened scrutiny and attention. The accuracy of every team’s “T-rank” now carries significant weight, adding pressure on Torvik to deliver. The prevalence of metrics in college basketball discussions has intensified, with terms like NET, KenPom, and Torvik becoming commonplace in evaluating teams ahead of Selection Sunday.
The NCAA Tournament selection process is undergoing a transformation, introducing a new metric with ties to professional gambling this year. There’s optimism within the analytics community that this addition could enhance clarity in the often complex selection process. However, the creators of these mathematical formulas shaping college basketball’s postseason dialogue feel the discussion isn’t always accurate.
A recent virtual roundtable organized by the NCAA assembled Torvik, Ken Pomeroy, Kevin Pauga, Matt Morris, and Alok Pattani, representing the seven metrics employed by the NCAA tournament selection committee. They collectively acknowledged the NCAA’s improvements in the selection process by discarding the Ratings Percentage Index (RPI). This diverse group, hailing from various backgrounds and locations, collaborates to offer insights into the meticulous task of selecting the 68 teams for the men’s NCAA basketball tournament.
Starting with the 2018-19 season, the NCAA implemented the NCAA Evaluation Tool (NET) and incorporated various rating systems. However, there was a consensus that only a select few of the seven metrics should be utilized in determining the 68 teams for the NCAA tournament. Notably, prominent figures like Pomeroy and Torvik advocated against using their rankings, while even the creator of the NET, Pattani, did not endorse its use in the selection process.
Despite the ambiguity surrounding the specific metrics to be employed by the selection committee, fans often focus on the NET rankings of teams during Selection Sunday, overlooking the intricacies of each rating system amidst the excitement of March Madness. These metrics are broadly categorized into predictive rankings, such as the NET, KenPom ratings, ESPN’s BPI, and Torvik ratings, which assess a team’s overall performance based on various factors, and results-based rankings like ESPN’s strength of record, KPI, and WAB, which evaluate a team’s achievements in relation to its resume.
The introduction of Torvik and WAB on NCAA team sheets has garnered interest, with Burn, the creator of WAB, suggesting that committee members could simplify their selection process by relying on this metric. Despite the confidentiality surrounding the exact formulas utilized for these metrics, there is a consensus that results-based metrics should play a significant role in team selection for the NCAA tournament.
While the selection committee remains tight-lipped about their approach to using these metrics, the NCAA stresses that the NET serves as a primary sorting tool rather than the sole determinant in the selection process. Ultimately, the human element of the 12 committee members collectively voting on at-large teams continues to be a crucial aspect of the selection procedure.
“During a recent teleconference, North Carolina athletic director Bubba Cunningham posed the questions: ‘Did you play, where did you play, how did you do?’ in his capacity as this year’s men’s NCAA tournament selection committee chairman. Cunningham also shared valuable advice from fellow committee members to consider team strength over metrics when evaluating potential selections. The NET ranking system, affiliated with the NCAA, garners significant public attention despite some criticisms. Bart Torvik, a lawyer from Evanston, Illinois, introduced his Torvik rankings in 2014, which will be included in team sheets for the NCAA Tournament selection committee starting in 2025. The evolution of bracketology metrics, notably the NET, was a result of collaborative efforts among various stakeholders in 2017. The outdated RPI system, based solely on winning percentages, gave way to the NET, which combines predictive and results-based elements in a single metric. Although initially met with skepticism, the NET has now become a crucial component in determining team strengths and quality wins in bracketology discussions. The NCAA’s adjustments to the NET formula for the 2020-21 season aimed to enhance its accuracy by focusing on statistics adjusted for efficiency. Despite differing opinions on the merits and potential flaws of the NET, the system remains a key tool for the selection committee in evaluating teams for the NCAA Tournament.”
Over the years, several entities including Torvik, Pauga, and ESPN have made adjustments to their formulas from their original versions. Unlike other predictive models, the NET does not incorporate preseason data or scoring margin, and treats every game equally regardless of when it was played. While the NET includes a team value index that is results-based, studies have revealed that the metric closely aligns with other predictive models. Clemson coach Brad Brownell had expressed his belief that Big 12 teams were able to manipulate the NET ratings by scheduling weaker nonconference opponents to inflate their efficiency numbers. However, the NCAA asserts that the NET’s correlation with KenPom, Torvik, and BPI indicates its effectiveness in fulfilling its purpose.
Despite the transition away from the RPI, some in the statistics community are still grappling with the creation, implementation, and shortcomings of the NET. Ken Pomeroy, the developer of the KenPom ratings, noted some similarities between his ratings and the NET, emphasizing the shift towards offensive and defensive efficiency in the formula. Since its inception in 2002, the KenPom ratings have been instrumental in advancing the use of advanced metrics in college basketball and March Madness.
Critics like Morris have labeled the NET as “inferior” due to its exclusion of scoring margin, suggesting it may not be the best metric for betting purposes. Despite its imperfections, ESPN bracketologist Joe Lunardi believes the selection process now heavily relies on metrics compared to when he began projecting NCAA Tournament brackets in 1995. Lunardi points out that while there may be discrepancies with the NET, cross-referencing with other popular metrics can help mitigate these inconsistencies.
The accessibility of data has propelled bracketology from a seasonal interest to a year-round industry. The Bracket Matrix, a website that monitors bracketology experts, has seen significant growth since its inception in 2006, now tracking over 90 bracketologists nationwide. This expansion reflects the public’s increasing interest in the bracket selection process and the utilization of various metrics to predict tournament outcomes.
Pauga, who also runs a business centered on a platform and algorithm named Faktor to aid conferences in scheduling, explained that the data-driven process is more structured and predictable, emphasizing that it does not intend to diminish traditional methods used by tournamentologists. According to him, the Selection Sunday surprise often revolves around the opponents faced rather than the teams being selected for the tournament.
The introduction of a new metric by the NCAA Tournament selection committee this year has sparked discussions among experts like Pomeroy, Torvik, and Worlock. This metric, Wins Above Bubble (WAB), assesses a team’s performance against its schedule compared to what a bubble team would be expected to achieve, utilizing the NET ranking system. Interestingly, it appears that this metric originated from a professional gambler named Seth Burn, who transitioned from accounting to successful sports betting due to his knack for analyzing data.
Burn’s post on his blog in 2015 introduced the WAB formula, which has now gained recognition among experts as a potential gold standard for team selection, free from human biases. The metric could level the playing field for teams with fewer high-caliber opponents, offering a more objective approach to determining the NCAA Tournament field. However, the ultimate decision rests with the selection committee, which may face challenges in justifying selections that contradict the WAB rankings.
The ongoing debate over selection metrics in the NCAA Tournament highlights the evolving landscape of team evaluation, with some advocating for a unified metric while Burn believes WAB could be the answer. As the tournament unfolds, the decisions made by the committee will be closely scrutinized, potentially shaping the future of team selection processes.
