ST. LOUIS POST-DISPATCH
Joe Martinich was as stunned as most viewers around here a year
ago, when he listened to Jim Nantz and Billy Packer accuse the NCAA
Tournament selection committee of a bias favoring mid-major teams.
The announcers "went overboard," Martinich said. "And, as we know, they
turned out to be wrong," because of the success of Bradley, Wichita
State and George Mason.
The mid-majors' success caused Martinich to ponder all the injustices
the committee is alleged to have perpetrated over the years. A
professor of operations management at the University of Missouri-St.
Louis, Martinich possesses the statistical tools and skills to
determine whether the committee had developed a long-term conference
bias.
First, he defined the majors as the six BCS conferences and the defunct
Big 8, all others as non-major conferences. Then, he chose to focus on
teams eligible for an at-large berth with an RPI between 21 and 79, the
range associated with bubble teams.
Martinich used the "gory details" from Jerry Palm's College RPI
data base, the same stats that the selection committee uses to pick the
tournament field. He focused on six independent variables: RPI rank;
conference RPI rank; net conference wins, determined by subtracting
losses to conference teams from victories over conference teams; total
wins against RPI top 25 teams, net wins against RPI 26-50 teams; and
net wins against RPI 51-100 teams.
The results of Martinich's statistical analysis matched the committee's
selections about 90 percent of the time. On average, two teams that fit
the statistical model were not selected by the committee, and two teams
that the committee selected did not fit the statistical model. Of the
31 discrepancies since 1994, 22 were within five places of the
selection cutoff point.
"When they say that RPI is just one tool, they mean it," Martinich
said. "The conference games and conference tournaments matter a lot."
The selection committee is quoted ad nauseam to that effect each year.
Martinich's statistics back their assertions.
Since the committee's work matches his statistical analysis so closely,
Martinich concluded that the committee appears to base its selection on
defensible, objective criteria and is relatively consistent, which
removes the threat of a consistent bias toward major or non-major
conferences.
But among the anomalies can be found vindication for Missouri State
fans. The Bears are the victims of the committees' two worst
injustices, in 2000 and this year. Martinich's analysis listed the
Bears 12 spots above the cutoff both years, yet the committee passed
them over.
Another anomaly appeared in 2005, when the NCAA changed the RPI formula
to give more weight to a road loss and less to a home win. From 1994 to
2004, the committee left out 10 teams from major conferences, 13 from
non-majors that fit the statistical model. Since 2005, though, the
committee has left out just one team from a major conference, seven
from non-majors.
So, if anything, the opposite of Packer's assertion has been true. In
technical terms, Martinich said, "Something's fishy. It almost looks
like they're basing their decisions on the old RPI formula, even though
they don't. This is something I want to look at."
Though he joked that he hoped his research would "predict future
outcomes so I could make a million and retire early," Martinich said
the NCAA could use it as a way to monitor selection and seeding. The
committee could use the research to defend their selections, since the
majority of decisions that bracketologists, fans and coaches hail as
injustices in reality are consistent.
His study "is a lot like instant replay," Martinich said. "When the
committee gets lambasted for being inconsistent, they can point to the
data and say, 'No. This is consistent. Look at our history.'"
