In 29 of the past 35 tournaments (1985 through 2019), one or more teams seeded 13, 14 or 15 have won a game in the Round of 64. Our research on causal inference using observational data ([1],[2]) has been adapted to look at key statistics that may predict upsets for these seeds. From 2003 through 2019, we used the Balance Optimization Subset Selection model (BOSS) to select two teams each year. Of the resulting 34 teams selected (out of a possible 204 teams that could have been selected), 11 ended up being upsets. During this time period, there were a total of 26 upsets. Below is a table, constructed using the BOSS model, which contains the two teams selected each years (from 2003 to 2019) that share the characteristics of previous upsets. We also report the actual upsets that occurred during this time period. Note that in 2018, No. 16 UMBC upset No. 1 Virginia. However, our method does not focus on No, 16 upsets.
| Year | Upsets Chosen by BOSS | Actual Upsets |
|---|---|---|
| 2019 | (13) St. Louis, (15) Abilene Christian (15) | UC-Irvine (13) |
| 2018 | (13)Buffalo, (14)Wright State | (13)Buffalo, (13)Marshall |
| 2017 | (13)Vermont, (14)New Mexico State | None |
| 2016 | (13)Iona, (14)Buffalo | (13)Hawaii, (14)Stephen F. Austin, (15)Middle Tennesse State |
| 2015 | (14)Georgia State, (14)UAB | (14)Georgia State, (14)UAB |
| 2014 | (13)New Mexico State, (13)Delaware | (14)Mercer |
| 2013 | (13)Lasalle, (14)Davidson | (13)Lasalle, (14)Harvard, (15)FGCU |
| 2012 | (14)Belmont, (15)Lehigh | (13)Ohio, (15)Lehigh, (15)Norfolk State |
| 2011 | (15)Akron, (15)LIU-Brooklyn | (13)Morehead State |
| 2010 | (14)Ohio, (15)Morgan State | (13)Murray State, (14)Ohio |
| 2009 | (13)Cleveland State, (13)Portland State | (13)Cleveland State |
| 2008 | (13)Sienna, (13)Oral Roberts | (13)Siena, (13)San Diego |
| 2007 | (13)Albany, (13)Holy Cross | None |
| 2006 | (13)Bradley, (14)Xavier | (13)Bradley, (14)Northwestern State |
| 2005 | (13)Vermont, (14)Niagara | (13)Vermont, (14)Bucknell |
| 2004 | (13)UI-Chicago, (13)Texas - El Paso | None | 2003 | (13)Tulsa, (14)Troy | (13)Tulsa |
Note that with the introduction of NET, which affects team seeds, we have not tested our algorithm with such information. As such, we will no longer report upsets for the first round.
References:
[1] Nikolaev, A.G., Jacobson, S.H., Cho, W.K.T., Sauppe, J.J., Sewell, E.C., 2013 “Balance Optimization Subset Selection (BOSS): An Alternative Approach for Causal Inference with Observational Data,” Operations Research, 61(2), 398-412.
[2] Sauppe, J. J., S. H. Jacobson, and E. C. Sewell, 2014, “Complexity and Approximation Results for the Balance Optimization Subset Selection Model for Causal Inference in Observational Studies,” INFORMS Journal on Computing, 26(3), 547–566.
For comments and feedback on this website, send an email to bracketodds@illinois.edu.
Follow us on twitter: @shjanalytics
Media requests should be made by either emailing bracketodds@illinois.edu
or calling 217-244-7275.
bracketodds.cs.illinois.edu © 2019 - The Board of Trustees of the University of Illinois.
All Rights Reserved.
