Statistical Modelling 7 (2007), 301313
Quantile regression for rating teams
Gilbert W Bassett Jr
Department of Finance,
University of Illinois at Chicago,
601 South Morgan (MC168),
Chicago, Illinois 60607–7121
U.S.A.
eMail:
gib@uic.edu
Abstract:
Quantile regression is proposed for modeling game out comes and as
the basis for rating teams. The model includes the standard location
model for team strength as a special case, while allowing for a
richer specification in which teams differ according to the quantiles
of the out come distribution. Team ratings are defined as the
handicap needed to equalize the out come of a contest. With teams
differing by quantiles, this leads to a class of ratings that
depend on where in the out come distribution the out come is
equalized. Relation ships with betting games are discussed.
The approach is illustrated by rating National Football League
(NFL) teams based on game results for the 2005 season.
Keywords:
handicaps; odds; pointspreads; quantile regression; sportsratings
Downloads:
Stata data and log files in
zipped
archive.
back