Over- And Underseeded Teams In March Madness Based On Predictive Rank

Selection Sunday always brings controversy. Each year, the NCAA selection committee combines resume-based results with predictive analytics to construct rankings, creating an inevitable gap between a team’s seeding and its actual strength. This year, resume-based metrics dominated seeding decisions. Predictive indicators quietly proved their worth in March. Last year, the bracket built around KenPom’s projections was around the 97.5th percentile, highlighting how efficiency-based predictive models better capture a team’s true strength than traditional resume evaluations.

This analysis isolates the prediction signal. Consensus rankings for team strength are derived by averaging three widely used efficiency metrics: KenPom, Bart Torvik T-Rank, and the ESPN Basketball Power Index (BPI). Comparing that ranking to the committee’s official order shows where the rankings align with underlying performance and where they diverge, revealing which teams are most likely to be overrated or overlooked heading into the tournament. These differences are where bracketed values ​​often appear.

advertise

March Madness Monitoring and Seeding Methodology

To evaluate how well March Madness seeding matches team strength, we constructed a data set that combines the committee’s full 68-team rankings with three widely used predictors: KenPom, Bart Torvik T-Rank, and the ESPN Basketball Power Index (BPI). These systems are designed to measure a team’s strength based on possession of the ball, regardless of wins or losses. Each team’s rankings across the three systems are averaged to create a single composite measure of predictive strength. Because all three metrics are expressed in ranking form, where lower values ​​indicate stronger teams, the average preserves that ordering while eliminating differences between individual systems.

The teams are then reranked from 1 to 68 based on this average, producing a “predicted ranking” that reflects what the field would look like if seeds were ranked based solely on underlying performance rather than resume. This ranking serves as a benchmark against which the committee’s official rankings are compared.

advertise

Calculate the difference between each team’s predicted ranking and committee ranking. A positive variance indicates that the team is monitored by the committee, while a negative variance indicates that the team is underseeded relative to its predictive power. The results highlight a clear divide between resume-driven selection and underlying team strength.

The consequences of overseeding and underseeding during March Madness

March Madness Seeds

The committee ranked some teams significantly higher than their projections indicated.

• TCU: TCU shows the largest gaps in oversight. The Horned Frogs moved up to No. 34 according to resume-based metrics. Prediction systems like KenPom and T-Rank rate them closer to the bubble at No. 42, suggesting their seeding may overstate their potential strength.

• Miami (Fla.): Miami’s ranking reflects strong resume metrics, but its efficiency metrics are more modest. The committee ranks them 27th in the field, while predictive indicators rank them 34th.

• UCF: UCF benefits from a resume that goes beyond its predictive capabilities. While the Cavaliers finished well, their efficiency metric showed the team ranked overall (No. 45) lower than their seeding (No. 38).

• UNC: North Carolina’s brand and resume performed well during the selection process, but the prognosticators were less promising. The gap between the committee’s No. 22 ranking and the metrics ranking of No. 29 suggests the Tar Heels may be seeded closer to the ceiling of their resume than their underlying performance.

March Madness Seeds

On the other hand, the predictive capabilities of some teams appear to be underestimated.

advertise

• Iowa State: Iowa State is one of the most underrated teams. Despite strong efficiency numbers, the Hawkeyes’ resume isn’t up to top standards, resulting in a lower ranking. The committee ranked it at 36th, compared with 27th, using predictive indicators.

• Louisville: Louisville’s predictive metrics suggest this team is much better than its seeding. The committee ranked Louisville No. 23, but the prognosticators had it ranked No. 14. This ranking does not take into account Mikel Brown Jr.’s injury.

• Tennessee: Tennessee combines strong efficiency metrics with solid but not elite resume metrics. That gap suggests the Volunteers may be one of the more dangerous teams relative to their seeding. They rank 21st by the committee and 13th in the predictive metrics.

What this means for your March Madness bracket

If predictive metrics are a reliable measure of team strength, then the gap between those rankings and committee seeding creates opportunities. Brackets are based on the resume. Predictive indicators are based on future performance. Teams with efficiency metrics above seed may be undervalued, while teams promoted via resume may be overvalued relative to their true strength.

In effect, this creates a form of bracket arbitrage. The market, in this case the committee’s seed, reflects a perception of the team’s quality. Predictive indicators offer another approach. When these perspectives do not align, there is an opportunity to exploit the gap. This does not guarantee results. March Madness remains highly volatile, and the single-elimination format introduces a level of randomness that no model can fully capture. But if predictive metrics like KenPom and Bart Torvik T-Rank are directionally accurate, teams identified as underseeded represent some of the most logical places to look for upside.