We examined the performance of the Bradley-Terry Model as a ranking tool for players and teams. We studied the effect of different tournament designs via the Fisher information of the parameters estimated using maximum likelihood. From the data and the theoretical analysis, we conclude that for the Bradley-Terry model to be of any use, the design of the tournament must be either highly connected or contains a large number of paired comparisons relative to the number of players.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Matthews, Gregory Jay

Publisher

• Worcester Polytechnic Institute

Identifier

• 03A005M

Advisor

• Morales, Carlos J.

Year

• 2003

Date created

• 2003-01-01

Resource type

• Major Qualifying Project

Major

• Actuarial Mathematics

Rights statement

• In Copyright