Answers are grounded in this paper and cite their sources.
Statistical modeling reveals improved agreement with expert polls for NCAA Division 1-A football rankings using win-loss data alone, highlighting a margin-independent evaluation strategy.
This article introduces a penalized maximum likelihood approach for ranking all NCAA Division 1-A college football teams. The model does not consider margin of victory and is based solely on win/loss data. Despite the simplicity, the model leads to rankings which exhibit greater agreement with expert opinion (i.e., AP and Coaches polls) than the models historically and currently used by the Bowl Championship Series (BCS). The model can be implemented using standard statistical software packages.
No takes yet. Share an insight, caveat, or question.
David Mease (2003) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: