Home Bracketology NCAA Selection Committee Will Show How New Elements Impact Decision-Making

NCAA Selection Committee Will Show How New Elements Impact Decision-Making

by Ian

This weekend, the NCAA Selection Committee will release their Top 16 teams for the NCAA Tournament. This does not mean that these will be the 16 teams on the top 4 seed lines come March, but it will provide an interesting look into the Committee’s thinking and decision-making process. I detailed earlier this year in my Bracketology 101 post the means and methods the Committee uses to determine the 68 teams in the Tournament field. They will conduct a mock exercise this week, only determining their Top 16 teams rather than the full field of 68. Last year was the first time this exercise was conducted, spurred by the success of the College Football Playoff Committee’s weekly rankings reveal.

Tean Sheet Changes

There are two changes this season that bracketologists like myself are excited to see how they will impact the Selection Committee. Both changes were brought about due to meetings between members of the media and the analytics community and the Committee. The first change was an adjustment to the Team Sheets that the Committee uses when comparing team resumes for the Torunament. The old Team Sheets broke down games based on opponent RPI rankings into four simple columns. Games against the Top 50 were in the first column, teams ranked 50-100 in the second, 100-200 in the third, and 200+ in the fourth. Here is an example of Providence’s actual Team Sheet from last year, which got them into the Tournament as one of the Last Four In.

This year, based on feedback from the analytics community, the columns in the Team Sheets have been slightly adjusted. Research has shown that winning a true road game against an inferior opponent is actually much more difficult than winning a home game against a better opponent. To account for that difference in game difficulty, the first column (now known as “Group 1” or “Tier 1”) contains home games against teams ranked in the Top 30, road games against teams ranked in the Top 75, and neutral court games against teams ranked in the Top 50. The essence of “Tier 1” has remained the same – the relative difficulty of playing a Top 50 team on a neutral court, but the contents have been adjusted based on statistical analysis that playing a Top 75 team on the road is the equivalent difficulty of playing a Top 50 team on a neutral court or a Top 30 team at home. Here is an example of how Providence’s Team Sheet would have looked last year if the new layout had been in place.

Incorporation of Other Metrics

The second change to the information available to the Selection Committee this year is the addition of other metrics to their analysis. For many years, members of the media have been taking shots at the RPI (the mathematical formula used by the NCAA to rate the quality of a team). Other metrics have been developed to rate the quality of a team, such as ESPN’s BPI that they try to force into everyone’s face but wasn’t something the Committee actually used. One of the hesitations of the Committee with rankings like the BPI and Ken Pomeroy’s rankings is that they involve a predictive element and are not purely based on results like the RPI. Now, the Committee has added six different mathematical calculations of team quality to their analysis. The metrics are divided up into 3 “Results-Based” rankings – the NCAA’s RPI (Ratings Percentage Index), the KPI (Kevin Pauga Index), and SOR (Strength of Record). There are also 3 “Predictive-Based” Rankings – ESPN’s BPI (Basketball Power Index), Ken Pomeroy’s rankings (POM), and Jeff Sagarin’s rankings (SAG). The Committee will likely use the average of the Results-based metrics, the average of the Predictive-based metrics, or the overall average of all 6 metrics. Currently, there are a few teams that are favored by one set of metrics over the other, and the initial reveal of the Top 16 will give us a look into how much the Committee values each of these new data points.

Predicting the Top 16

The Top 16 will be officially revealed on Sunday. For now, here is my best guess at how the Top 16 will fall.

#1 Seeds: Virginia, Villanova, Xavier, Purdue

With Villanova’s loss to St. John’s, they no longer hold a claim to the top overall spot in the rankings. Virginia, with a 23-1 record and 13 wins in Tier 1 and 2 combined (tied for the most in the nation) has the clear claim to the #1 overall seed. Villanova is still deserving of a spot on the top line as they are tied for the most Tier 1 victories with 9 and have very good computer numbers across the board. The interesting case comes with the next three teams. Xavier has 12 wins in Tiers 1 and 2 and no bad losses. Xavier ranks well in the results-based metrics but does not fare so well with the predictive metrics. If the Musketeers are not on the 1-seed line on Sunday, we will know the Committee is weighting the predictive metrics heavily. Purdue fell to a good Ohio State team this week, and the question remains whether they should stay on the 1-seed line or be bumped by Kansas. The Jayhawks have more good wins than the Boilermakers (3 in the Top 25, 9 Tier 1 wins vs 1 in the Top 25 and 5 Tier 1 wins) but Kansas also has 3 losses outside of Tier 1 while Purdue only has 1. I chose Purdue as my last #1 seed, but the margin was very close.

#2 Seeds: Kansas, Tennessee, Texas Tech, Clemson

I debated for quite some time the merits of teams with five Tier 1 wins vs the teams with four. At this point, the next 7 teams all have between 9 and 11 wins in Tiers 1 and 2 combined. The difference is in the number of good wins and bad losses. Kansas was an easy choice because of their overwhelming number of good wins. Ultimately, I chose these teams because of their wins against top competition. Tennessee has 4 wins against Top 25 opponents while Texas Tech and Clemson both have 3. Neither Tennessee or Clemson has dropped a game outside of Tier 1 and Texas Tech has only lost one.

#3 Seeds: Auburn, Cincinnati, North Carolina, West Virginia

As opposed to the #2 seeds, first two #3 seeds on my list have fewer good wins. While all of the #2 seeds had at least 3 wins against Top 25 opponents, Auburn and Cincinnati have 1 and 0 respectively.  The Tigers and Bearcats have both avoided bad losses, which keeps them high on the list. Cincinnati is another interesting case for the predictive metrics. Those metrics love the Bearcats (an average of 4.3 in that category) but they are not as favored by the Results-Based metrics. Cincinnati’s positioning in the Top 16 will be another indicator of how the Committee feels about the new metrics. North Carolina, fresh off their big win over Duke, has more good wins than any team remaining (4 vs the Top 25, 5 vs Tier 1) but also has two bad losses that bring their profile down. West Virginia has one fewer top-end win but also has not lost to a Tier 3 team. I gave WVU the edge over Duke (despite Duke being favored by the computer rankings) due to the Mountaineers having more good victories on their resume than the Blue Devils.

#4 Seeds: Duke, Oklahoma, Ohio St, Michigan St

This may be low for both Duke and Michigan State, both of which have very good computer numbers across the board. However, Duke only has 7 victories in Tiers 1 and 2 combined while Michigan State has only 5. Partially due to the relative lack of quality of the Big Ten and also because of bad non-conference scheduling, Michigan State and Ohio State both have only 2 Tier 1 victories, the fewest of any team in the Top 16. Oklahoma is another team that the predictive metrics do not favor, but the Sooners have 6 Tier 1 victories and only 1 loss to a Tier 2 team, which is better than any other remaining team can boast.

Just Missed: Kentucky, Butler, Rhode Island, Texas A&M, Creighton, Arizona, Miami

In the end, I will probably be wrong in my predictions, but the rankings revealed by the Committee will be an interesting insight into how they will approach the final bracket in March. At this point in the season, it’s about learning how the Committee will handle the new information that is available to them and making future predictions based off of those tendencies.

You may also like

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.