On Sunday, the NCAA Tournament Selection Committee revealed their Top 16 seeds (as of now) for the NCAA Tournament. This was in part a practice exercise for the Committee to go through the motions of selecting, ordering, and bracketing a group of teams. Additionally, it gave the bracketology community at large some insight into how the Committee views the changes to team sheets and new results-based and predictive metrics. Here is the initial seed list produced by the Committee.
New Metrics Matter
Last week, I discussed how the Committee had added five computer-generated ranking metrics to the Team Sheets in addition to the RPI. Three of the metrics are purely “results-based.” The traditional RPI (Ratings Percentage Index), which is a formula used by the NCAA across multiple sports, the Kevin Pauga Index (KPI), and a Strength of Record (SOR) calculation. Three additional metrics are “predictive-based” and include ESPN’s BPI (Basketball Power Index), Ken Pomeroy’s rankings (POM), and Jeff Sagarin’s rankings (SAG). The Team Sheets also include an average of the results-based metrics and predictive-based metrics (we’ll call them RBM and PBM for these purposes) as well as an overall average ranking of the 6 metrics. As with most things in bracketology, an individual number does not tell you much, but the comparison of that number to other numbers provides some insight into the quality of a team.
Of the 16 teams that appeared atop the seed list, the Top 14 all had an overall metric average lower than 16. Only Arizona (21.00) and Oklahoma (23.83) were above 16. None of the teams that were left out had an overall average under 16. The best overall averages left out were West Virginia (17.83), Wichita St (19.33), Gonzaga (19.67), and Rhode Island (20.50). Oklahoma (the “last team in”) had more Tier 1 victories than all of these teams, which goes to show that good wins are still weighted heavily by the Committee. It should be noted that Oklahoma is favored by the results-based metrics has a fairly wide disparity between their RBM (17.33) and PBM (30.33). Of the teams with the best averages that were left out of the Top 16, Gonzaga, West Virginia, and Wichita St all fared very well in the predictive metrics but were ranked lower in the in the results-based metrics. Conversely, Rhode Island scored very well in the results-based categories while not faring as well with the predictive metrics.
While the new metrics did matter, the Committee did not solely rank teams based on those averages. Xavier was seeded ahead of Purdue despite a lower metric average (and a much lower predictive-based average). Auburn was seeded 5th overall, above 6 teams with better metric averages. Clemson was seeded above three teams with better metric averages. That being said, Duke was slotted in the #7 overall spot, despite having fewer Tier 1 wins (4) than 3 teams seeded behind them and fewer Tier 1 and 2 wins (8) than 6 teams seeded behind them. Interestingly, Duke was seeded higher than Clemson, who also has 4 Tier 1 wins, has more Tier 1+2 wins (10 to 8), but has a lower metric average (8.00 for Duke to 12.17 for Clemson).
Tier 2 Losses Are Forgivable
Of the teams in the Top 16, 11 of them have lost at least one game to a team outside of Tier 1. Like with losses to teams ranked 50-100 in the past, the Committee did not heavily dock teams who suffered Tier 2 losses. Three of the four #1 seeds (Virginia, Villanova, Purdue) lost one of their games to a Tier 2 opponent. Virginia and Villanova were seeded higher than Xavier, who had no losses outside of Tier 1. Duke received a #2 seed despite a Tier 2 loss as well as tying for the fewest Tier 1+2 wins of anyone in the Top 16. Michigan State was slotted on the 3-seed line despite a Tier 2 loss and only 3 Tier 1 wins (only Ohio State had fewer amongst the Top 16). The last two teams in the Top 16 (Arizona and Oklahoma) both had 3 Tier 2 losses along with lower metric averages than teams like Gonzaga and Rhode Island but made the Top 16. This leads me to conclude that Tier 2 losses are forgiveable and Arizona/Oklahoma’s 3 losses in that quadrant are not seen as decidedly worse than Gonzaga’s 1 or Rhode Island’s 0 (or at least, not enough to drop their seeding and offset their quality wins).
Tier 3 Losses Are Deadly
While the Committee appeared to be lenient with Tier 2 losses, they did penalize teams for suffering Tier 3 losses. Tier 3 is the relative equivalent of sub-100 teams in the old system (and Tier 4 is sub-200 teams). Only 3 teams in the Top 16 had a loss in Tier 3 on their resumes – Kansas, North Carolina, and Ohio State. Those bad losses against Tier 3 opponents (Kansas losing at home to Oklahoma St, UNC losing at home to Wofford, Ohio State losing at home to Penn St), definitely impacted Kansas and North Carolina’s seedings. Kansas has the best win portfolio in the nation – 9 Tier 1 wins and 14 Tier 1+2 wins, both marks the highest of any team. However, they were slotted on the 2-seed line and given the #6 overall seed, indicating that Auburn with no losses outside of Tier 1 was closer to the top line than the Jayhawks. Head-to-head, it would seem that Kansas profiles as a superior team to Auburn. Kansas has 9 Tier 1 wins (Auburn 4), 14 Tier 1+2 wins (Auburn 9), and a metric average of 8.50 (Auburn 10.17). However, Auburn with no losses outside of Tier 1 was slotted ahead of Kansas who has a Tier 2 and a Tier 3 loss.
Similarly, North Carolina was seeded 12th overall, the last of the #3 seeds. UNC has 7 Tier 1 wins (the most of any team ranked lower than 6th) and 10 Tier 1+2 wins (second only to Texas Tech’s 11 of the teams on the 3 and 4-seed lines). However, UNC was seeded below Duke (7th), Clemson (9th), and Michigan State (11th). UNC has more Tier 1 and Tier 1+2 wins than Duke and beat the Blue Devils head-to-head. Duke has a slightly better metric average and did not lose to any Tier 3 opponents. Clemson also has fewer quality wins than UNC (4 Tier 1 wins vs 7), a worse metric average (12.17 for Clemson vs 9.84 for UNC), and split head-to-head with the Tar Heels. However, the Tigers have not lost a game outside of Tier 1 and were seeded higher than UNC. Finally, Michigan State beat UNC head-to-head, has a similar metric average, and has fewer quality wins (3 Tier 1 wins vs 7, 8 Tier 1+2 wins vs 10). However, once again, the Committee clearly docked North Carolina for their Tier 3 loss and moved them below a team whose resume is decidedly inferior.
Here is a look at the overall resumes of the Top 16 teams the Selection Committee included in the bracket: