Up to this point, the findings of the study have not proven that Euclidean distance is a relevant predictive factor for NCAA Tournament game outcomes. While this is somewhat disappointing, there is yet reason to suspect that spatial factors may be proven relevant if they are designed somewhat differently. However, currently, the authors must conclude that the analysis to date should most certainly not become required study for bettorsor even the casual "bracketologist"!
In reviewing the results, it is clear that RPI and seed are by far more effective means of prediction, especially when considering the relative differences in these values between the teams involved. These factors will continue to be the basis for any study of the tournament, but further consideration is warranted as to how spatial factors might enhance the predictive analysis to more accurately anticipate upsets and outcomes.
In the discussion period after the authors' presentation of a paper on the study at the 2006 Esri International User Conference, the following suggestions for factors to include were made:
- Route distance, as opposed to Euclidean distance
- Air travel versus road travel
- Number of games played
- Number of overtime games played
- Regular season winning percentage
- Pod seeding (i.e., brackets)
These factors will be considered in further study research, although intuition suggests that the most important next step will be to evaluate the criteria against betting lines. While they hold no inclination toward improving bookies' capabilities or even their own ability to beat the odds, the authors suspect that the impact of distance traveled may be more relevant to the actual score differential in games, rather than to the simple win/loss outcomes. For example, a significant distance or even home-court advantage is not going to allow 16 seed Fairfield to beat national powerhouse Indiana in the first round. In fact, no 16th seeded team has ever won a first-round game in the NCAA Tournament. However, spatial factors may be relevant in determining whether Fairfield University is trounced by 34, or competes before losing by a dozen.
Regardless of the authors' further investigations into spatial indicators of NCAA Men's Basketball Tournament success, they will undoubtedly be tuned in next Marchas they are every yeartrying to figure out which 12 seed will upset a 5 seed this year.
The authors thank Bryce Wells for his creative input into the original study and for mastering its Web site (www.geohoops.com), Shannon McElvaney and Barb Schmitz for their moral support, and their fantastic academic mentorsDawn Wright of Oregon State and Steve Steinberg of Humboldt Statewho continue to inspire them to think outside the box.
About the Authors
Brian Ward is a GIS developer and analyst with CH2M HILL, Inc., in Colorado Springs, Colorado. He holds a bachelor's degree in professional geography from the University of North Alabama and a master's degree in geography from Oregon State University. His research interests include the analysis and visualization of spatial/temporal phenomena. He can be reached at 719-477-4917 or firstname.lastname@example.org.
Brian Davenhall is a GIS developer and analyst with CH2M HILL, Inc., in Redding, California. He holds an associate's degree in geographics from Hocking College and a bachelor's degree in natural resources planning from Humboldt State University. He is well versed in mobile application development and integration and is a Trimble Certified MGIS Trainer. His research interests include interpolation and imputation methods for data estimation, geostatistical modeling, and sampling design. He can be reached at 530-229-3253 or email@example.com.