WAB nailed March Madness
Thayer Partners argues the ranking system that best ‘nailed’ the 2026 March Madness champion was WAB (Wins Above Bubble), making the point that the most useful model depends on the bracket task, not raw ranking accuracy (Thayer Partners). That’s a useful reminder for analysts — optimize models for the decision you need (bracket success) rather than a single accuracy metric (Thayer Partners).
The funny part of this year’s men’s bracket is that the metric built to judge résumés, not predict point spreads, lined up with the team that cut down the nets. Michigan finished No. 1 in the NCAA’s Wins Above Bubble table through games on April 6, and Michigan then beat Connecticut 69-63 for the 2026 national title on April 6 in Indianapolis. (ncaa.com 1) (ncaa.com 2) Wins Above Bubble means this: take a team’s schedule, ask how an average “last-team-in” tournament team would have done against it, and compare that baseline with what the team actually did. The NCAA says if a bubble team would have won 19 games against a schedule and Team A won 20, Team A gets a plus-1.0 Wins Above Bubble score. (ncaa.org) That makes Wins Above Bubble a résumé tool, not a crystal ball. It ignores margin of victory and efficiency and focuses on results, the same way a hiring manager might care more that you closed the deal than whether your presentation software loved your slides. (ncaa.org) (si.com) The NCAA has been leaning harder on that kind of résumé lens in men’s basketball. In its March 5, 2025 explainer, the NCAA said the committee uses the NCAA Evaluation Tool as its primary sorting tool, but also listed Wins Against Bubble as a new men’s metric used in 2024-25 to evaluate what teams actually accomplished. (ncaa.org) That distinction is the whole point of the Thayer Partners piece. Their argument is not that Wins Above Bubble is the single “best” ranking system in every sense; it is that the best system depends on the job, and the job in a bracket pool is to identify the kind of team profile that survives six games in March. (thayerpartnersllc.com) Michigan is a clean example of why that can work. The Wolverines were No. 1 in the NCAA’s Wins Above Bubble ranking at 18.11, ahead of Arizona at 16.26 and Duke at 15.01, and they reached the title game as a No. 1 seed before winning the championship. (ncaa.com) (sports-reference.com) Connecticut shows the same idea from the other side. The Huskies were No. 4 in Wins Above Bubble at 11.93 and still made the title game out of the East, which is what you would expect from a résumé metric that is trying to capture who has stacked strong results against a hard schedule. (ncaa.com) (sports-reference.com) The trap for analysts is treating one accuracy score like a universal scoreboard. A model that is great at saying Team X would beat Team Y on a neutral floor is solving a different problem from a model meant to mimic committee logic or flag teams built to survive a single-elimination bracket. (ncaa.org) (thayerpartnersllc.com) College basketball has been full of acronym churn for years, from the Ratings Percentage Index to the NCAA Evaluation Tool to Key Performance Indicator, and Sports Illustrated noted in February 2026 that Wins Above Bubble had become the latest metric fans were being told to learn before Selection Sunday. The reason is simple: the committee’s decisions are not made from one number, so the smartest bracket work starts by matching the number to the decision. (si.com) (ncaa.org)
