By Henry Johnson
March Madness is here, and with it are high hopes and lost revenue. Last night, the selection committee released the bracket for the 2017 tournament, and the ensuing uproar was enough to drown out even the rowdiest student sections.
Much of the fury was caused by the supposed underseeding of the Wichita State. The Shockers were “jobbed” and “screwed” and the whole thing was downright “criminal.” College basketball fans are up in arms, but if it’s any consolation, the numbers are on their side. Wichita State got a raw deal.
To look at a team’s deserved seed vs. its actual seed, I used Ken Pomeroy’s adjusted efficiency margin and ranked the tournament teams according to this metric. This ranking was compared to the selection committee’s ranking of the 68 teams in the field.
Unsurprisingly, there’s an overwhelmingly strong relationship between these two measures. The pressing question is which teams received seeds that were far higher or lower than Pomeroy’s metric suggested was proper. To answer this question, we can subtract the difference between a team’s Kenpom ranking (within the field of 68) and its committee ranking.
Sure enough, Wichita State is this year’s most underseeded team, with a whopping 30-spot difference between its deserved and actual ranking. According to kenpom, the Shockers should have been as high as a 2-seed, a far cry from the 10-seed they find themselves in.
Maryland is the most overseeded team of the tournament according to kenpom. The Terrapins scored a 6-seed despite having an adjusted efficiency margin worse than lesser-seeded conference peers like Wisconsin, Michigan, and Northwestern.
Perhaps Wichita State was the victim of a committee that’s skeptical of mid-majors. To try to detect the role of conferences in seed determination, I built a model that predicts a team’s overall tournament seed (1-68) based on its Kenpom efficiency margin. I then added an indicator variable for mid-majors. In defining mid-major, I did what any college student would do and defaulted to Wikipedia.
One quick technical note is that this method is far from perfect. Besides taking a team’s kenpom score as gospel, it’s somewhat flawed statistically since overall seed is a ranked variable rather than a continuous one.
When we run a regression, the resulting equation is as follows:
NCAA Overall Seed = -1.658*kenpom + 7.47*Mid-major + 58.8
The mid-major variable is statistically significant, with a p-value of 0.005. The model suggests that, controlling for a team’s efficiency margin, we expect a mid-major to be seeded 7.47 spots lower than its high-major and power conference counterparts. If we assume every 4 spots in the overall ranking corresponds to a 1 seed jump in the bracket, the model suggests that mid-majors are about 2 seeds lower than deserved on average.
For the technical reasons discussed earlier, we should be a bit careful about this interpretation. After all, this method could project decimal-value seeds or even negative seeds.
Still, using ordinal regression, we see the same significant relationship. This general finding falls in line with past research on the topic. For more information on bias among power conference teams, check out this excellent post by HSAC’s own Kurt Bullard.
As for Wichita State, despite being a 10-seed, FiveThirtyEight is calling them a favorite against Dayton in the Round of 64. There may be no better karmic retribution against the committee than a deep Wichita State run. We’ll have to see whether Ball Don’t Lie when it comes to tournament seeds.