This year’s bracket reveal went mostly according to expectations. The biggest surprise was definitely Virginia getting in. I did not see that coming at all, and I am struggling to figure out why they got in.
One interesting data point is from bracketmatrix.com. If you’re not familiar with that site, it’s an compilation of hundreds of bracket predictions. It includes all the well-known ones from Joe Lunardi, Jerry Palm, SI.com, USA Today, FOX Sports, etc., but it also includes predictions from people like me. Of the 200 bracket predictions on that site, only 20 of them had Virginia in. 135 brackets had Oklahoma in; 74 had St. John’s; 22 had Indiana State. St. John’s and Indiana State were my first two out.
Why did Virginia make it? There are essentially two possible answers: politics or resume. Perhaps they made it because Tony Bennett is well-connected, or perhaps the committee felt sorry for the ACC, or perhaps the Cavaliers enlisted Tom Sheehey to make selection committee chairman Charles McClelland an offer he couldn’t refuse… I don’t have any inside information about whether any of that is true, and I’m not going to speculate on it. For the sake of this post, I’m going to assume that the Cavaliers’ resume had something that the committee liked better than Indiana State, Oklahoma, and St. John’s. What was it?
The only thing that stands out in that regard is that Virginia ranked higher in what they call the Results-Based Metrics. There are five different “computer rankings” (if you’ll excuse the loose use of that term) that the committee looks at. Two of them are results-based, meaning their rankings look backward at what a team has done. You might think of these as the resume metrics. Three of them are predictive, meaning their rankings are forward-looking, i.e. what they expect a team to do in the future.
The predictive metrics are familiar to many fans: kenpom, Jeff Sagarin, and the ESPN BPI. The results-based metrics are not as well known. One of them is called the Kevin Pauga Index (KPI), and the other is called Strength of Record (SOR).
I don’t have a clear picture on how these metrics are calculated. Strength of Record is described as the probability that an average Top 25 team would have the team’s record or better, given the schedule. OK, I can kind of understand that intuitively, even if I don’t know exactly how it’s calculated. The Kevin Pauga Index is proprietary. According to ncaa.com:
KPI ranks every team’s wins and losses on a positive-to-negative scale, where the worst-possible loss receives a value of roughly around -1.0 and the best-possible win receives a value of roughly 1.0. KPI then averages these scores across a season to give a score to a team’s winning percentage. The formula uses opponent’s winning percentage, opponent’s strength of schedule, scoring margin, pace of game, location, and opponent’s KPI ranking.
Alrighty then… in any case, whatever these metrics are, they favor Virginia. Here is how the bubble teams compare in these metrics:
| Team | KPI | SOR | Average |
| Virginia | 38 | 32 | 35 |
| Oklahoma | 50 | 28 | 39 |
| Indiana State | 40 | 40 | 40 |
| Colorado State | 28 | 56 | 42 |
| Pitt | 55 | 48 | 51.5 |
| St. John’s | 71 | 45 | 58 |
I don’t pretend to have enough information to know what is going on here, but I will engage in a little speculation. One of the many ways to evaluate a team is, what was their record, and what was their strength of schedule. If Team A and Team B have a similar strength of schedule, and Team A has a better record, then it follows that Team A is more deserving, right?
That reasoning is simplistic, but there is a certain persuasiveness to it. So I decided to look at Virginia compared to other teams with a similar strength of schedule. Of course I don’t know exactly how these ratings calculate strength of schedule, so I will use kenpom’s. Here is a selected group of teams with similar strength of schedule and record to Virginia:
| Team | SoS Rank | W-L |
| Colorado State | 65 | 24-10 |
| Kentucky | 66 | 23-9 |
| Oregon | 67 | 23-11 |
| Nebraska | 68 | 23-10 |
| Duke | 75 | 24-8 |
| Colorado | 76 | 24-10 |
| Virginia | 77 | 23-10 |
| Washington State | 84 | 24-9 |
Note, every team in this group made the tournament. And Virginia seems to fit right in. They are almost identical to Colorado. They compare reasonably well to Washington State (one game worse record, but tougher schedule).
Again, this is speculation on my part. But best I can tell, these resume metrics, especially Strength of Record, are doing something like this. Virginia did beat Florida and Texas A&M, and they did go 13-7 in the ACC – better than Pitt, Wake, Clemson, or NC State.
That’s the only good thing about Virginia’s resume that I can find. Their Quadrant 1 record was 2-7, so that wasn’t it. They look terrible on the predictive metrics, and that’s because, well, getting blown out is very bad for your predictive metrics. They lost to Wisconsin by 24, Memphis by 23, Notre Dame by 22, NC State by 16, Wake by 19, Virginia Tech by 34, and Duke by 25.
Oklahoma in particular has reason to complain. Do you realize, they did not lose a single game below Quad 1? There were undefeated in Quad 2/3/4 games. They are the only team in recent memory to go undefeated against Quad 2/3/4 and fail to get an at-large bid. Perhaps they were dinged for their non-conference strength of schedule, which was ranked only 262. But that’s deceptive; they actually played Iowa, USC, Providence, Arkansas, and UNC and went 4-1 in those games. It’s not their fault that USC and Arkansas turned out to be bad this year. What really caused that 262 ranking is that they also played four of the worst teams in Division I – Mississippi Valley State, Texas-Rio Grande Valley, Arkansas Pine Bluff, and Central Arkansas. Who cares? They played five major conference teams and a bunch of cupcakes. That’s a completely normal schedule. Should it matter that their cupcakes were even softer than everyone else’s?
Look, I’m not crying for these other teams. Their resumes were deeply flawed as well, and these are fine distinctions that have to be drawn. I personally am an advocate of emphasizing the resume more and the predictive metrics less. But the idea that some guy Kevin Pauga is the reason that Virginia is in and St. John’s is out… can you blame the coaches for being frustrated with that?