Upsets and March Madness

The last decade of men’s college basketball has seen increased parity between the power conferences and mid-major conferences.  Friday’s NCAA Tournament results epitomize this shift.  Prior to Friday number 2 seeds won 96.4% of their games against number 15 seeds over the past 27 years.  Both number 15 seeds defeated their second seeded opponents when Lehigh upset Duke and Norfolk State upset Missouri on Friday.  These upsets would have been even more surprising had they occurred in the 1980’s and 1990’s when the gap between college basketballs’ haves and have-nots was much wider.  In the past decade the NCAA’s pod system has given more favorable site assignments to highly seeded teams and limited the number of upsets mitigating the impact of increased parity.

 Today the best players often stay for a single year of college basketball before turning pro.  Consequently the top seeds are usually less experienced and start more freshman than many of the lower seeded teams.  This was not true in the early years of the 64 team format, when teams from traditional powers such as Kansas, Kentucky, North Carolina, and Duke had more talented players and more experienced teams.

March Madness upsets are now more likely to occur because the gap between top and lower seeded teams is the smallest since the 64 team format was introduced in 1985.  One way to measure the gap between stronger and weaker teams is through the Ratings Percentage Index, or RPI.  Teams with a higher RPI have defeated better teams that played stronger schedules during the regular season. 

The RPI methodology has changed over the years, so I compare relative RPI differentials in first round tournament games.  The smallest RPI differentials are in games between teams seeded 8 and 9, and the next smallest differentials are in games between teams seeded 7 and 10.  I normalize the double difference in RPI in opening games involving 7-10 and 8-9 seeds as a baseline for measuring other RPI differentials.

During the 1980s and 1990s the relative RPI differential for the average number 5 seed was 6.6 times the baseline RPI differential.  Over the past 6 tournaments the average number 5seed faced a relative RPI differential that was only 1.5 times the baseline differential.  The same is true for other highly seeded teams.  During the 1980s and 1990s the RPI differential for the average 2seed was 24.4 times the baseline RPI differential, and has dropped to only 8.3 times the baseline differential in the past 6 tournaments.

This means that today’s number 12 seeds are similar in strength to number 10 seeds in the 1980s and 1990s, and number 15 seeds are similar to previous number 12 seeds.  As I mentioned in an earlier post, my research with Todd McFall indicates that the NCAA’s pod system has mitigated the effects of increased parity by giving more favorable site assignments to the top four seeds in each region.  Despite what we observed Friday night, more favorable site assignments for the top 4 seeds in each region has reduced the number of upsets during the first weekend of March Madness.

%d bloggers like this: