Metrics For Each Seed In March Madness

This year’s Selection Sunday drama captivated college basketball fans nationwide. The unveiling of the NCAA Tournament field of 68 is more than just a television spectacle, however. It is the culmination of a complex and often contentious process of team evaluation using various statistical metrics. Each system provides a unique perspective on which teams deserve a shot at March Madness glory. In 2025, the selection committee used two new metrics for the very first time: the Torvik T-Rank and the WAB (wins above bubble). These were used in addition to five other metrics including NET, KenPom, BPI, Strength of Record (SOR), and KPI rankings. In general, SOR, KPI, and WAB are considered resume-based metrics, while NET, KenPom, BPI, and Torvik are considered prediction-based metrics. I used statistical methods to determine which metric most closely aligned with the committee’s seeding decisions.

ForbesNCAA Tournament 2025: The Seeds Most Likely To Break Your March Madness BracketBy Giovanni Malloy

Ranking Comparison Methodology

Data

I gathered data on the full seed ranking of the 2025 NCAA men’s basketball tournament, as well as the final NET, KenPom, BPI, Torvik, SOR, KPI, and WAB rankings. By compiling this comprehensive dataset, I compared each metric’s final rankings against the committee’s seeding decisions. Importantly, I adjusted the rankings so that each metric was scaled from 1 to 68, reflecting each team’s relative standing solely within the tournament field rather than their original national rankings. I also did the same with the average ranking among all seven metrics, the average ranking among resume-based metrics, and the average ranking among prediction-based metrics.

This comparison will help identify which metrics were most closely aligned with the committee’s choices and potentially reveal biases or preferences in the selection process. Furthermore, analyzing how well each ranking system correlated with seeding will allow us to uncover whether resume-based metrics (SOR, KPI, WAB) or prediction-based metrics (NET, KenPom, BPI, Torvik) played a more significant role in shaping the 2025 tournament field.

Measuring Accuracy

To identify which ranking system most closely matched the NCAA Selection Committee’s decisions, I used three different accuracy metrics: Spearman’s Rank Correlation, Kendall’s Tau, and Mean Average Precision (MAP).

Spearman’s Rank Correlation is a way to measure how similar two rankings are, even if they are not a perfect match. If two lists rank teams in nearly the same order, they will have a high Spearman score. It is designed to spot overall trends and patterns.

Kendall’s Tau is more about consistency in paired comparisons. It checks how often one system ranks a team higher than another compared to what the committee did. The more often they agree, the better the score.

MAP focuses on precision in the most important spots. It rewards ranking systems that do the best job at identifying the top teams correctly.

Using all three metrics provides a well-rounded look at how closely each ranking system matched the committee’s seeding decisions.

Most Accurate Rankings

The table below shows how well each ranking system’s order of the 68 tournament teams matched the NCAA Selection Committee’s actual seeding. For Spearman and Kendall, higher scores indicate better alignment with the committee’s choices. For MAP, lower scores indicate better alignment.

The results were telling. The Spearman’s Rank Correlation scores showed how closely each ranking system’s overall order of teams matched the committee’s seeding. Among all systems, WAB (0.98) had the highest Spearman score, suggesting that its ranking most accurately reflected the committee’s perception of overall team strength. Prediction-based metrics trailed behind resume-based metrics.

When considering Kendall’s Tau, which focuses more on consistency between two ranking systems, taking an average of all ranking metrics had the highest accuracy (.91). Among individual metrics, WAB (0.90) again led the pack. This score indicated a high level of agreement on how individual teams were placed relative to one another. Interestingly, resume-based metrics again outperformed prediction-based metrics.

The MAP metric focuses on identifying the most important teams accurately. Lower scores are better, and once again, taking an average of all ranking metrics performed best (1.02). Among individual metrics, WAB (1.03) and SOR (1.03) demonstrated the strongest precision in identifying the top-seeded teams. Even for top seeds, the resume-based metrics were more accurate than prediction-based metrics.

Takeaway

The data signal the NCAA Selection Committee’s slight preference toward resume-based metrics over prediction-based metrics when making seeding decisions. On average, SOR, KPI, and WAB all performed better than NET, KenPom, BPI, and Torvik across all three metrics. This suggests that the committee places a higher value on the actual results teams produce rather than pure predictive power. Most notably, WAB emerged as the most influential metric on the committee’s seeding decisions, outperforming all others in terms of agreement with the final seed rankings for this year’s March Madness.