Academic journal article
By Stern, Hal S.
The American Statistician , Vol. 58, No. 3
1. ON SPORTS AND CHAMPIONSHIPS
Professional sports leagues in the United States tend to be organized as collections of small (four to eight teams) divisions that play unbalanced schedules (more games against opponents within the division than against other teams). This promotes strong local rivalries but implies that one cannot just compare records at the end of the season to determine a champion. Instead the regular season is used mainly to identify participants for a post-season championship tournament. It is the post-season tournament's champion (e.g., the World Series winner in baseball and the Super Bowl winner in football) that is universally recognized as that season's champion. Division or conference titles earned during the year are celebrated but not with the same level of enthusiasm as the championship.
This is of course not the only model for professional sports. For example, football (soccer) leagues in Europe are generally organized as 18- to 20-team leagues that play a balanced regular season schedule composed of home-and-away games with each team in the league. The regular season championship in such leagues is highly valued. There are also tournaments and competitions, both intranational (e.g., the FA Cup in England) and international (e.g., Europe's Champions League), considered as separate events. The tournaments resemble the American post-season tournaments. Tournament titles are valued but not necessarily more than the regular season title.
American college sports (except football) have evolved a kind of compromise of the two models. University teams are organized in athletic conferences. The conference regular season title is usually based on a balanced schedule, though this is not always true now that conferences have grown in size. Regular season conference champions along with some of the best nonchampions are invited to a post-season national championship tournament. As in professional sports the post-season championship tournaments have become the most sought-after title, the one that is highly recognized. Unlike the professionals it seems that regular season conference championships remain highly valued.
College football's most competitive level (Division I-A) is unique among American sports. There is no post-season championship tournament. (There are championship tournaments in the football divisions for smaller schools). Instead for many years the college football national champion has been crowned based on the final votes of two polls (the Associated Press poll of sportswriters and the USA Today/ESPN poll of coaches). In recognition of the fact that the championship is settled by polls rather than on the field, the college football title has long been known as the "mythical" national championship.
The remainder of this article considers the way in which college football currently selects its national champion and how ideas from the field of statistics might improve the process.
2. AMERICAN COLLEGE FOOTBALL--PAST AND PRESENT
It is natural to start with a brief history of the search for a college football national champion.
2.1 The Early Days
A first key point is that there has long been interest in knowing which college football team is the nation's best. This is not a modern phenomenon. The earliest recognized rating system designed to address the topic was developed by Frank Dickinson of the University of Illinois in 1924. Teams were awarded points for winning games, with the number of points awarded determined by the quality of the opponent. Dickinson stopped producing his ratings in 1940, citing his age, a declining interest in college football, and the large number of alternative systems that had been developed (Dickinson 1941). Thus, by 1940 there is evidence of great national interest in identifying the best college football team in the land.
2.2 The Rise of the Polls
In 1936 the Associate Press (AP) news agency began a national poll of sportswriters to identify the top teams. …