The NCAA men’s basketball selection committee is working diligently to assemble and seed the teams that will participate in March Madness and be announced on Selection Sunday, March 17. However, the protocol that the committee is using is broken. The results for the 2023 men’s tournament provide some anecdotal evidence, with a disproportionate number of head-scratching upsets that may be indicative that seedings may not have accurately reflected team performance.
For the first time in the modern era, since 1986, no teams seeded No. 1, 2 or 3 reached the Final Four. Nor did any No. 1 seed reach the Elite Eight. At the same time, the sum of the Final Four seeds was 23, its second highest ever.
When the selection committee moved from the rating percentage index, or RPI, to the NCAA evaluation tool, or NET, this was a positive move forward to capture team performance. Incorporating the KenPom metric was a welcome addition. However, by weighting all games during the season the same, the committee is incorrectly seeding teams, resulting in numerous upsets that are becoming more commonplace.
This is reflected in how the end-of-season NET ranking may not reflect a team’s performance earlier in the season. For example, a pedestrian win in November may look like a mind-blowing victory on paper in March. The dynamic nature of NET means that the numbers of Quad 1 and 2 wins or Quad 3 and 4 losses constantly change, though the actual outcome of all such games is fixed in time.
This is, of course, how the NBA regular season standings are determined, with all wins equally weighted in team records. However, their playoffs are best-of-seven series, not single game eliminations.
The most blatant example of this phenomenon in 2023 was Purdue University, which was given a No. 1 seed. It started 22-1, including Top 25 victories over Marquette and Gonzaga. The team’s season ended at the hands of Fairleigh Dickinson University during the Round of 64. Though this outcome was certainly a surprise, it was not shocking, given how Purdue performed over the prior month. Seeding it as a No. 1 seed may have been deserving based on its NET metrics, but misrepresented its performance at the end of the season, as its trajectory was certainly not reflective of a top seed.
The same was the case with Iowa State and Illinois, both of which had marquee early season wins but limped to the finish line and were quickly eliminated in the Round of 64.
If the selection committee wants to have the very best teams awarded at-large bids, the definition of “best” must be rethought. This can be achieved by weighting games not only by which teams played but by when they played.
Early season games rarely provide an accurate assessment of a team’s performance in March. Since the selection committee encourages scheduling preconference games against the very best competition, a November win or loss gets weighted into at-large bid assessment and seeding the same as a late-conference outcome.
This is misguided and results in misseedings, rewarding and penalizing teams inappropriately.
One solution is that conferences begin their conference play earlier in the season, allowing time for nonconference games in February. Though such a change would require coordination across multiple conferences, it would provide a more accurate assessment of teams. The Big 12/SEC Challenge, held in late January 2023, provided a model for such competition and scheduling. Unfortunately, it has come to an end.
Upsets happen. It is a natural process in any sport. That is why games are played on the court and not on a computer. If the likelihood of a team winning a game is 90%, this means that if the game was repeated 100 times, then on average, the better team would win around 90 of these games.
March Madness is a single elimination tournament, which is unforgiving to any team that may be having an off day against an inferior opponent that is playing at its peak. That is why the overall No. 1 seed in the tournament is not anointed the national champion before the tournament begins. That team has to earn the title by winning six games, typically against teams that it has not played during the season, including low-majors and mid-majors, in a style that it may not be accustomed to.
Given that the 2024 selection committee will likely overseed teams that began the season strong but limped to the finish line, and underseed teams that found their mojo in February and March, the phenomenon of irrational upsets will likely continue. Predicting them specifically with certainty will always remain difficult.
So when you pull up your app and catch some games March 21-22, remember that just because a team has been given a seed by the selection committee, the criteria used may obfuscate its true abilities and how it will perform once the opening tip has been made.
Sheldon H. Jacobson, Ph.D., is a professor in computer science at the University of Illinois at Urbana-Champaign. He is also the founder of the Bracketodds website, a science, technology, engineering and mathematics learning lab at the university.