For a better browsing experience, please upgrade your browser.

FiveThirtyEight

Sports

Louisville is FiveThirtyEight’s pick to win the NCAA tournament even though it’s a No. 4 seed. That doesn’t reflect a bluegrass bias, or a weak draw, or a bug in the model (we hope). Instead, it reflects how massively underseeded the Cardinals are. According to our calculations, Louisville is the seventh-most underseeded team in the past 12 tournaments, and the most aggrieved this year.

That’s the good news for Cardinals fans, whose team has a 15 percent chance of winning the title, according to our model. Louisville is also the plurality pick for Midwest Region champion of ESPN, Yahoo and CBS readers who picked brackets. It even has the backing of President Obama, who has the defending champion Cardinals returning to the final (though losing to Michigan State).

Now here’s the bad news: Although teams that receive rough treatment from the selection committee outperform their seed on average, there are plenty of exceptions. The most underseeded team among top-four seeds since 2003 — Memphis in 2009, a No. 2 seed rated by some analysts as the strongest team in the country — lost in the Sweet Sixteen.

We can place Louisville’s No. 4 seed in historical context in several ways, none ideal. I chose to calculate, for each team in the tournament since 20031, its expected winning percentage against the other teams that received that seed during that period, using ratings from college basketball analyst Ken Pomeroy. So, I’m pitting Louisville against the other 47 teams that got No. 4 seeds since 2003, in a hypothetical mini-season, to see how much stronger the Cardinals are than their peers.2

According to these calculations, Louisville would have won 69.8 percent of games against other No. 4 seeds, assuming strength in each season was the same. That’s the highest winning percentage relative to seed since 2003 for a No. 4 seed. At the other end of the spectrum, Vanderbilt, in 2008, would have beaten just 29.6 percent of fellow No. 4 seeds — the lowest of the group.

These calculations don’t necessarily mean the selection committee hasn’t done its job when seedings don’t perfectly reflect team strength. Some years, the gap between the four best teams and the next four — or eight or 12 — is greater than it is in other years. Also, sometimes rules such as preventing early meetings between conference rivals can force the committee’s hand. This season, the committee mostly got things right. Only one team other than Louisville is badly underseeded: Tennessee, the No. 11 seed in the Midwest.

However, the selection committee chairman, Ron Wellman, said that his panel continues to use Rating Percentage Index to evaluate whether brackets are balanced. RPI’s continued influence, despite its well-known drawbacks, suggests that some teams do get bad breaks.

One reason the committee might not feel it needs to change its seeding practices: Teams like Louisville haven’t always soared. Despite being stronger than their seeds suggest, many underseeded teams underperform.

Five of the six teams that were more underseeded than Louisville fell flat. Four lost their first games. (One of those, Lamar in 2012, lost a play-in game.) Memphis, in 2009, fell in the round of 16 despite its impressive stats. Only Florida last year, as a strong No. 3 seed, exceeded expectations, falling in the Elite Eight.

Outperformance of seed is a tricky concept. The simplest way to evaluate how teams should do is to assume they should beat all lower seeds, but that supposes that all No. 1 seeds will always make the Final Four, which is far from automatic. So for each team since 2003, I started with the number of games it won — from the round of 64 on — and subtracted from that the average number of wins for all other teams in that period with that seed.3 So, for instance, the average No. 2 seed since 2003, excluding Memphis in 2009, won 2.4 games. Memphis won two. So it underperformed by four-tenths of a game.

There’s a small, positive correlation4 between a team’s hypothetical winning percentage against teams with the same seed and a team’s wins relative to expected wins for its seed. So, on balance, teams that are underseeded do prove the selection committee wrong. But the effect is so minor, and clouded by so many exceptions, that they hardly add up to a compelling case against controversial seeding decisions. There are too many examples like Memphis in 2009 to set against cases such as Davidson in 2008, a No. 10 seed that would win two-thirds of games against other 10 seeds and that made the Final Four.

Team Year Seed Avg Games won Expected wins Over-performance
Utah St. 2005 14 77% 0 0.09 -0.09
Belmont 2011 13 71% 0 0.30 -0.30
Memphis 2009 2 70% 2 2.40 -0.40
Pittsburgh 2013 8 70% 0 0.67 -0.67
Florida 2013 3 70% 3 2.16 0.84
Louisville 2014 4 70% TBD 1.48 TBD
Virginia 2007 4 30% 1 1.49 -0.49
Gonzaga 2006 3 30% 2 2.19 -0.19
Vanderbilt 2008 4 30% 0 1.51 -1.51
UNC Asheville 2003 16 27% 0 0.00 0.00
Miss. Valley St. 2008 16 26% 0 0.00 0.00

Louisville can search its own media guide for examples of teams that should have done better than their seeds but didn’t. In 2004, the Cardinals were heavily underseeded at No. 10. That suggested they were between the 37th- and 40th-strongest team in the tournament, despite being ranked 16th in Pomeroy’s ratings. Louisville drew a Xavier team that was ranked lower by Pomeroy but that lost by 10. In 2011, like this year, Louisville got a No. 4 seed it could have complained about, because Pomeroy rated the Cardinals 11th and five teams rated lower got better seeds. Yet the Cardinals lost their first tournament game.

The potential for upsets that makes the NCAA tournament so exciting could also reduce the Cardinals quickly from title favorite to disappointment, which would end any debate over whether they were underseeded. Anyone lobbying the selection committee to reconsider its decisions should beware of staking the case on any one team, even one as strong and as underseeded as Louisville is. After all, even our model anointing it the favorite gives Louisville an 85 percent chance of not repeating as champions.

Footnotes

  1. As far back as our data set goes. ^
  2. To make it easier to compare teams across different years’ tournaments, I used Pomeroy’s Pythagorean rating for a team: his estimate of that team’s winning percentage if it played a schedule full of average teams. He bases it on teams’ schedules and their offensive and defensive efficiency — how many points per possession they score and allow. (Pomeroy provided, at my request, his Selection Sunday ratings for each year going back to 2003.)

    We can get win probability for a game using a relatively simple calculation involving the teams’ Pythagorean ratings, outlined here. Teams in different seasons aren’t usually compared in this way, because each season’s ratings are benchmarked to that season’s average team. In our case, we’re looking for a team’s strength relative to its peers in a given season — compared, in turn, to the strength relative to its peers of another team that got the same seed. So the calculation is appropriate.

    Comparing teams to just the other teams with their seeds in the same season produces similar results. Louisville would have won 70.6 percent of games in a hypothetical season against the other three No. 4 seeds this year: Michigan State, San Diego State and UCLA.

    Other ratings also take into account similar factors to Pomeroy’s; ESPN’s Basketball Power Index adds the feature of accounting for injuries but goes back just three seasons. Some other possession-based ratings rank Louisville even higher than Pomeroy does. That is reflected in FiveThirtyEight’s model, which gives Louisville a 15 percent chance of winning the title, compared to 12.3 percent by Pomeroy. So another rating might show Louisville to be even more underseeded than Pomeroy’s does. ^

  3. This is known as “performance against seed expectation” and has been calculated going back to the start of the current tournament format.  ^
  4. R=0.12 ^

Filed under , , , ,

Comments Add Comment

Never miss the best of FiveThirtyEight.

Subscribe to the FiveThirtyEight Newsletter
×

Sign up for our newsletters to keep up with our favorite articles, charts and regressions. We have two on offer: The first is a curated digest of the best of what’s run on FiveThirtyEight that week. The other is Ctrl + , our weekly look at the best data journalism from around the web. Enter your email below, and we’ll be in touch.




By clicking subscribe, you agree to the FanBridge Privacy Policy

Powered by WordPress.com VIP