Analysts Vary in Opinion of Harvard's Upset Chances

One of the most remarkable facets of the Crimson’s win over Cincinnati is that Harvard did not play like an underdog. The Bearcats was the favorite but was not beaten by a typical underdog pulling in offensive boards, creating turnovers and draining an abnormally high number of threes. Cincinnati was simply outplayed over the course of 40 minutes despite being the clearly more athletic team. This begs the very simple question: can Amaker’s boys do it again on Saturday against Michigan State?

Fortunately March is when the sports stats community comes out of hibernation in the name of ‘bracketology’, with multiple predictions trying to put a number on the Crimson’s chances.

ESPN's Giantkillers blog gives the Crimson at a lowly 24.0 percent of pulling out the upset, but acknowledge that they believe that to be a “lowball estimate”. ESPN points to Harvard playing actually a fairly average game in the second round with an uncharacteristic 12 turnovers and only nine offensive rebounds as evidence that the Spartans need to watch out as the Crimson can step up their game.

According to kenpom.com, Harvard has shot an excellent 38.8 percent from beyond the arc, good for 26th in Division-I, but only 28.5 percent of its field goals are three point attempts, in the bottom quintile nationwide. On Thursday, the Crimson upped its three point rate to 38.6 percent—reminiscent of its 42.8 percent rate last year against New Mexico—and will likely do so again against the Spartans.

Underdogs benefit from high variance tactics, and there is no higher variance shot than the trey. The Spartans rank a barely-above-average 112th in the country at allowing three pointers to drop. If Rivard and Curry can get going as they did in the first half against Cincinnati, Michigan State will be in danger.

Ken Pomeroy, a meteorologist and leader in tempo-adjusted college basketball analytics, uses the entire season of data to put a Pythagorean winning percentage on each college hoops team. Using the Pythagorean numbers from two opponents, we can estimate the probability of each team winning using Bill James’ log5 formula. Using this method, Harvard comes out a lot better with a 36.0% chance of getting to the Sweet Sixteen. However, Ken Pomeroy himself acknowledges that “there are subjective factors which this analysis does not include” when predicting Tournament games.

Nate Silver, of predicting elections fame, uses kenpom’s rankings along with a whole host of other predictive metrics in his analysis. A fuller explanation of his method is here but there are two key reasons why he is not as high the Crimson, giving them only a 23% chance to beat Michigan State. Firstly, he incorporates the preseason rankings which had the Spartans narrowly as second only to Kentucky, which he argues represents an estimation of the true talent on their roster independent of the relatively short 35-game season. Secondly, he utilizes Sonny Moore's ratings, which emphasize recent performance very heavily. Since Michigan State is one of the hottest teams in the country right now, with 22.3 percent of ESPN users picking them to make the Final Four—the most popular team from the East—it scores very highly via this metric.

Finally, if you want a model which likes Harvard, look no further than a recent Harvard graduate. John L. Ezekowitz ’13 produces a bracket annually based on survival analysis. His bracket analysis also makes unique use of mapping the network of Tourney teams based on regular season interactions, and then integrates network properties of each team into his estimations. John’s predictions have the Crimson winning 38 percent of the time.

Of course, outside of the statosphere is the betting line, which reflects public opinion on the game. As of press time Harvard is a +274 underdog, which translates to an expected winning percentage of 26.7 percent.

So while everyone disagrees on how much of a chance the Crimson have on winning tomorrow, everyone agrees it has a chance.

—Staff writer Julian Ryan can be reached at jryan01@college.harvard.edu.