College basketball season is in its final sprint toward Selection Sunday, and that means it’s panic time for teams on the men’s NCAA Tournament bubble. But not all bubble teams are created equal. For instance, Wake Forest, which ranks 30th in ESPN’s BPI power rankings, ought to feel relatively good about its tourney chances, while Syracuse, which ranks 31st, should be in a state of full-blown terror. Why the difference? It all comes down to how each school did in a metric that best predicts the NCAA selection committee’s picking tendencies.
The selection committee has struggled for a while with how to pick the teams for the NCAA tournament. Should they be the best teams? Or simply the most deserving teams? The former may not have the best record, while the latter may not have the best talent. (In January, the NCAA even met with several analytics experts to help sort out this quandary.)
Stats, too, have to figure out which of these two questions they’re trying to answer. ESPN’s BPI and KenPom.com’s ratings are examples of predictive rating systems, meant to tell fans who the “best” teams are. Other rating systems such as ESPN’s new Strength of Record (SOR) are meant to rank “deserving” teams, grading a team’s resume according to how hard its record was to achieve. The two approaches can yield very different results, and my research shows that Strength of Record aligns more closely with what the committee looks for in a tournament team.
As the man behind updating ESPN’s College Basketball Power Index (BPI) before this season, I studied the Selection Committee’s historical behavior, to help build a system that assigns each team a probability of making the NCAA Tournament. I looked at many variables, including resume rankings such as Strength of Record and the NCAA’s official Ratings Percentage Index, predictive rankings such as BPI and KenPom, and simpler measures such as a team’s win-loss record.13
Most of these variables are highly correlated with each other, and in many cases they tell the same story about a school’s season. But Strength of Record has the highest correlation, and using it alone to select teams would help you agree with the committee 90 percent of the time. Plugging Strength of Record into a model (which includes other metrics, but weights them less14), we can assign tournament probabilities to teams considered “on the bubble” by ESPN’s Joe Lunardi. Because of differences in resume quality (as measured by SOR), some bubble teams are quite likely to be favored by the committee, while some should be worried if they fail to win their conference tournaments.
| CURRENT RANKING AMONG D-I TEAMS | |||||
|---|---|---|---|---|---|
| TEAM | RPI | BPI | KENPOM | STRENGTH OF RECORD | CHANCE OF MAKING TOURNEY |
| Wake Forest | 34 | 30 | 30 | 33 | >99.9% |
| Xavier | 33 | 36 | 43 | 37 | 99.9 |
| Marquette | 55 | 29 | 28 | 41 | 88.6 |
| Illinois State | 31 | 60 | 49 | 46 | 85.7 |
| USC | 39 | 58 | 61 | 43 | 76.4 |
| Vanderbilt | 44 | 49 | 40 | 56 | 65.6 |
| Providence | 52 | 55 | 53 | 40 | 63.0 |
| Kansas State | 60 | 41 | 32 | 45 | 61.9 |
| Rhode Island | 42 | 39 | 51 | 54 | 58.1 |
| Indiana | 81 | 32 | 46 | 57 | 19.1 |
| Houston | 50 | 35 | 41 | 59 | 17.2 |
| Clemson | 63 | 33 | 35 | 49 | 15.7 |
| Syracuse | 79 | 31 | 44 | 53 | 12.2 |
| California | 56 | 52 | 55 | 69 | 2.8 |
| Illinois | 58 | 63 | 68 | 58 | 1.5 |
| Iowa | 72 | 77 | 67 | 64 | 0.5 |
So if your team is sweating things out on the bubble Sunday, keep an eye on its Strength of Record — not its RPI or its predictive ratings.