The first ranking of the 2019 College Football Playoff will be announced at 7 p.m. MST Nov. 5 on ESPN, but a bespoke algorithm developed by Joe Craig of UCCS and Niven Winchester of the Massachusetts Institute of Technology takes its own look at who the top 10 college football teams are in the NCAA’s top division.
Craig, associate professor of economics; and Winchester, principal research scientist at MIT and founder of RugbyVision.com; developed the algorithm to get people to take a harder look at their own biases, approach the ranking of college football teams in a more objective manner, and to illustrate how analytics can be used to inform decision making.
“Ranking systems have inertia,” Craig said. “It’s rare to see teams moving up drastically or moving down drastically unless there is a major upset. Moreover, rankings generated too early in the season create a bias where people continue to think that a highly-ranked team is good, and then if another team beats them, then that team must be good. Often times, both teams aren’t great but we don’t know until late in the season. As a result you end up with a team like Notre Dame in the playoff last year that had no business being involved in the College Football Playoff. We believe that taking an unbiased look at the performance of teams that applies no initial ranking – the rankings are based only on results from the current season – is important to use as a tool for deciding the College Football Playoff.”
The algorithm ranks teams by assigning points based on game scores after adjusting for the strength of the opposing team and accounting for home advantage. The points are the hypothetical points that a team would score in a neutral venue. So if Ohio State were to play Alabama, the formula would predict a Buckeye victory by 16 points in a neutral location.
The researchers said that the current rankings are based only on games played to this point of the season. Using Clemson as an example, currently ranked third in the coaches poll and fourth by the Associated Press but seventh in this ranking, they said that Clemson may not be the seventh-best team in the nation at the end of the season. But the algorithm indicates that they are at this point based on how they have preformed against the teams they’ve played
Other teams in these rankings that Craig expects to draw attention based on the differences with the main two rankings:
- No. 3 Wisconsin: The Badgers have two losses and are No. 16 in both the AP and coaches poll. But one of those losses was to top-ranked Ohio State and Wisconsin has large wins against Michigan and Michigan State, which both rank highly in the algorithm.
- No. 4 LSU: The Tigers are No. 1 in the AP poll and second in the coaches poll, any many pundits believe they have beaten three of the top 10 teams this season. But one of those teams, Texas, has dropped out of the AP and coaches poll and is only ranked No. 23 in the algorithm. The other two wins against Auburn and Florida, No. 10 and No. 13 in the algorithm respectively, were both at home and by the expected margins.
Looking ahead to the big games this weekend, Craig and Winchester provided their insight based on the full rankings:
- Alabama is expected to defeat LSU by five points.
- Penn State is expected to defeat Minnesota by 18 points.
- Wisconsin is expected to defeat Iowa by 17 points.