Is Harvard going back to the NCAA Tournament?

By Julian Ryan

Harvard basketball was ranked #25 in the AP preseason poll, marking the first top-25 preseason ranking in school history. The last time an Ivy program was ranked in the preseason was Penn in 1975. The ranking reflects the team’s fantastic wins in the NCAA tournament over the past two seasons and is a momentous achievement for the program (okay, homer rant over).

But what does the ranking mean, or more broadly: what information does the preseason AP poll give us about teams’ expectations? Nate Silver still uses the preseason rankings in his March Madness model as even after the regular season transpires, preseason projections retain significant predictive value

To analyze this, I collected all the AP preseason rankings and votes going back to the 2002-2003 season and then saw how they did in the NCAA tournament at season’s end. I was interested primarily in whether teams qualified for the tournament, and secondly on how well they did if they qualified. The results are broken down by groups:

Preseason AP Rank Average Tournament Wins Made Tournament? Sweet Sixteen Elite Eight Final Four






























As expected, most of the teams ranked in the preseason made it to the Big Dance (82.8% for the whole sample) and the higher ranked teams do better, with the top five teams making it to the Elite Eight about half the time. Surprisingly though, teams in the bottom tier largely outperformed teams in the second to last tier, both making the Tournament more often and winning more Tournament games on average. This anomaly requires further research but one hypothesis is that some bad major conference teams with a prestigious history are the recipients of positive voter bias, at the expense of less well known mid-majors who end up either ranked lower or not at all. This year UConn, Michigan State and Ohio State are ranked #17, #18 and #20 respectively and while all three have been final four threats over the past decade, perhaps they are a little overhyped this year.

So this rough estimate puts Harvard’s chances of making it back to the tournament at 74.2% but we can do better than that. After all, Harvard is ranked 25th and only received 98 votes (the lowest in my entire sample). To produce a better rough estimate, I made logistic regression models using a team’s ranking and votes to estimate Harvard’s chances of Tournament qualification, and of advancing to the second weekend and beyond:

Make Tournament?


Make Sweet 16?


Make Elite 8?


Make Final Four?


I omitted teams’ conference affiliation from the model because it is difficult to know where Harvard stands in this regard. It is definitely harder to make the Tournament as a good team from a one-bid rather than a multi-bid conference as you have to win your end-of-season tournament which has much higher variance than an at-large bid from a season’s worth of a resume.  However, Harvard is unique in that it plays in the only one-bid conference without a postseason tournament, with the Ivy League regular season champ gaining the automatic spot. My intuition would be that Harvard’s chances of qualification are slightly higher than the number posted above, given the relative (knock on wood) ease with which Harvard should roll over the Ancient Eight.

Beyond that, Harvard is still ultimately just a good mid-major and dreams of a 2009-Cornell-esque Cinderella story seem like a long shot. The team should take heart though from the 2003-04 Oklahoma State team. Ranked #25 with an eerily similar 102 voting points in the preseason, the Cowboys rode the junior-senior dynamo of Joey Graham and Tony Allen (read Siyani Chambers and Wes Saunders) all the way to the Final Four. Stranger things have happened.

About the author


View all posts

1 Comment

  • Interesting analysis, Julian. You touched on this at the end of your post, but if a team from a power conference underachieves significantly respective to their preseason rank, e.g. UCLA this year, they face long odds to earn their conference’s automatic bid (by winning the PAC-12 tournament); whereas, a mid-major team that underachieves respective to their preseason ranking, e.g. Harvard this year, may still remain in close contention for an automatic bid, and thus may have a greater chance to reach the tournament than your original model predicted. One thing you might consider is adding preseason conference strength to your model. Sticking to the preseason poll, you would need to use a surrogate variable (or two) to measure conference strength. I would suggest using the number of teams from the same conference in the top 25 and a teams rank respective to other conference teams in top 25. I imagine these variables would add predictive value – particularly in the case of a team like Harvard that has less competition for an automatic qualifying bid. As you said, this still doesn’t take into account the lack of a conference tournament in the Ivy League, which would be hard to capture given the lack of data (i.e. Ivy league teams ranked in the preseason top 25 over the years).

Leave a Reply

Your email address will not be published. Required fields are marked *