At the start of last season’s NCAA tournament, many pundits predicted No. 13 seed Eastern Washington would upset No. 4 seed Georgetown in the first round of March Madness.
“You got Georgetown sitting in there, they’re dying to be upset based on the way they played in the Big East tournament,” analyst Doug Gottlieb said on CBS’s “Selection Sunday,” which announced last year’s bracket March 15.
Eastern Washington Head Coach Jim Hayford guaranteed his team would upset the higher-seeded Georgetown.
“We’re going to win and talk again, Jim,” Hayford said on CBS Sports Radio’s “Jim Rome Show” prior to the game.
Part of the reason that Eastern Washington generated so much chatter as an upset pick was that Georgetown has been susceptible to upsets in the past. Current NBA superstar Stephen Curry burst onto the national scene in 2008 by scoring 30 points in No. 10 seed Davidson’s upset of No. 2 seed Georgetown. In 2013, No. 15 seed Florida Gulf Coast became only the seventh No. 15 seed to defeat a No. 2 seed when it upset Georgetown. FGCU earned the nickname “Dunk City,” as it became a national sensation on its way to the Sweet 16.
However, last year, Georgetown silenced its critics by defeating Eastern Washington 84-74. Although the Hoyas lost to No. 5 seed Utah in the next round, Georgetown was finally able to avoid the upset bug in last year’s tournament for the first time since it lost to No. 1 seed Ohio State in the 2007 Final Four, having lost to No. 10 Davidson, No. 14 Ohio, No. 11 VCU, No. 11 North Carolina State and No. 15 FGCU in that span. As the Hoyas finally managed to avoid an early upset in last year’s tournament, it is an appropriate time to analyze the statistical probability of NCAA tournament upsets.
The first step in this process is to determine what exactly constitutes an upset. For the sake of simplicity, any game that includes a team that is at least three seeds higher than its opponent would be deemed a potential upset. A No. 3 seed versus a No. 6 seed would be considered a possible upset, but a No. 4 seed beating a No. 2 seed would not qualify. Over the past three years, 155 games fit this criteria. The higher seed won 72 percent of these matchups.
Analyst Ken Pomeroy’s website kenpom.com compares teams on an objective scale, using series of statistics, including points scored and allowed per possession, strength of schedule and the number of possessions a team is likely to have per game. This gives a numbers-based idea of how teams stack up without relying on the highly fallible eye test. Ratings began in the 2002 season, making it a perfect tool for hindsight.
Using a logistic model that picks the result of the game, the likelihood of an upset occurring can be calculated. The logistic model accurately predicted whether the underdog would win the game around 80 percent of the time.
According to the model, Georgetown was expected to beat Eastern Washington 92.99 percent of the time. Fortunately for the Hoyas, the outcome that played out on the court fell in the 92.99 percent.
That was not the case in 2013 or 2010. According to the model, Georgetown had a 93.06 percent chance of beating FGCU before suffering a shocking upset. In addition, No. 3 seed Georgetown had a 92.15 percent chance of beating No. 14 seed Ohio in 2010, but lost 97-83. However, according to the model, the Curry-led Davidson team was an underdog in terms of seeding only, as the Hoyas had only a 20.48 percent chance of defeating the Wildcats.
The influence, or lack thereof, of pace of play will interest critics of Georgetown’s Princeton-style offense. As it turns out, the tempo of a game had little impact on the outcome. Even a side-by-side comparison of teams that won or lost and their respective tempos revealed a minimal difference between the two groups. Thus, it probably is not fair to blame upset losses on the Princeton offense’s tempo. In any event, Head Coach John Thompson III believes that this year’s team can play at different paces.
“I think we can be a fast, smaller team, I think we can jump. If we have to, I think we can win different ways,” Thompson said.
Anything can happen in the Big Dance, and no amount of predictive modeling can decide the outcome of a game ahead of time. However, models like the one used above indicate that despite past history, Georgetown is likely to avoid upsets in the future.