TY  - JOUR
AU  - Shen, Gang
A2  - Shen, Gang
A2  - Hua, Su
A2  - Zhang, Xiao
A2  - Mu, Yingfei
A2  - Magel, Rhonda
DB  - BISp
DP  - BISp
KW  - Analyse
KW  - Analyse, statistische
KW  - Basketball
KW  - Kondition
KW  - Leistungsprognose
KW  - Mannschaftssport
KW  - Methode, statistische
KW  - Prognose
KW  - Prognosemodell
KW  - Ratingskala
KW  - Sportspiel
KW  - Sportspielforschung
KW  - Sportwissenschaft
KW  - Statistik
KW  - Trainingswissenschaft
KW  - Untersuchung, vergleichende
LA  - eng
TI  - Predicting Results of March Madness Using the Probability Self-Consistent Method
TT  - Vorhersage der Resultate des March Madness mithilfe einer selbstständigen Methode
PY  - 2015
N2  - This paper proposes a new bracketing tool for all the 63 games in the last 6 rounds of the NCAA Men’s Basketball Tournament. Our new method is based on a binomial generalized linear regression model with Cauchy link on the conditional probability of a team winning a game given its rival team. It uses publicly available team ratings and game statistics prior to the March Madness. The new method is compared to three existing methods currently used to help complete March Madness brackets. The three existing methods include bracketing based on RPI ratings (RPI, 2014); bracketing based on Pomeroy ratings (Pomeroy, 2014) which use the Pythoagrean winning expectation; and bracketing based on estimated probabilities from a model developed using the restricted OLRE method proposed by West (2006, 2008). It is expected that our method will do better than West’s method since the OLRE method does not take into account which teams actually play each other. It is also expected that the proposed method will do better than using Pomeroys’s ratings or RPI ratings since Pomeroy’s ratings have done better than RPI ratings (Pomeroy, 2014) and our method also uses the Pythgorean winning expectation in which Pomeroy’s ratings are based. Nine years of March Madness data from 2002 to 2010 is used to develop models for both the OLRE method and our new proposed method. The four methods are tested using data from March Madness 2012, 2013, and 2014. Overall, our method did better than the other three methods in predicting March Madness winners, particularly when the double scoring system is used. Verf.-Referat
L2  - http://www.sapub.org/global/showpaperpdf.aspx?doi=10.5923/j.sports.20150504.04
L2  - https://dx.doi.org/10.5923/j.sports.20150504.04
DO  - 10.5923/j.sports.20150504.04
SP  - S. 139-144
SN  - 2169-8759
JO  - International journal of sports science
IS  - 4
VL  - 5
M3  - Elektronische Ressource (online)
M3  - Gedruckte Ressource
ID  - PU201508006285
ER  -