Preseason college football polls certainly have their flaws; generally the same programs end up on top, with the most recent bowl results holding a disproportionate amount of weight in the eyes of voters.
ESPN Stats & Information created preseason FPI to take the subjectivity out of the process. It uses data that has been found to be predictive of future performance to project how strong each FBS team will be in the coming season.
In 2014, the first season of FPI’s preseason ratings, the top three teams in preseason FPI –- Florida State, Oregon and Alabama -– made the College Football Playoff. Last season was harder to predict, but every playoff team ranked in the top 20 in the preseason. For more on FPI’s performance last year, read here.
Heading to the 2016 season, LSU sits atop the first iteration of preseason FPI (there will be multiple updates throughout the offseason), but to understand the Tigers’ projection it is important to make sense of what goes into the ratings.
What goes into preseason FPI and why?
Preseason FPI comprises four major components, which, based on statistical modeling, interact with one another to produce a team’s rating.
The first and most important is prior years’ performance on offense, defense and special teams, with the most recent year counting most. Opponent-adjusted expected points added (EPA), the backbone of all of ESPN’s college football metrics, is utilized to measure performance in prior seasons.
It makes intuitive sense that the most recent year’s performance generally has the greatest impact, but previous years’ performances are also predictive. Alabama has ranked in the top 12 in defensive EPA in each of the last four seasons, including being No. 1 in 2015. Despite losing six defensive starters, it’s no surprise that the Crimson Tide have the top defensive projection of 2016.
The second-most important factor is returning starters, which is provided by ESPN Insider Phil Steele. All teams –- even bad ones –- are helped by added experience, so returning starters on offense and defense are included in the preseason ratings. Because of the way starter information interacts with other components of the preseason ratings, each added starter is worth a different value for every team. For example, returning 10 offensive starters to an efficient offense would have more added value than returning 10 starters to an inefficient one.
Nonetheless, returning starters have value, and returning a starting quarterback has more value. Preseason FPI also incorporates information on transfer quarterbacks, an ever-growing trend, so teams that add a transfer quarterback with at least a season’s worth of action plays receive partial credit for having a "returner" at that position.
In 2016, LSU returns an FBS-high 18 starters (nine on offense and nine on defense), including running back Leonard Fournette and quarterback Brandon Harris, which is a major reason the Tigers sit atop preseason FPI 1.0. Similarly, Florida State returns all 11 offensive starters to a unit that has ranked in the top 25 of efficiency in each of the last four years. As a result, the Seminoles have the top offensive projection heading to 2016.
The final two pieces of preseason FPI hold significantly less weight but are important when predicting team strength. FPI utilizes a four-year average recruiting ranking of four systems (ESPN, Scouts, Rivals and Phil Steele) to measure talent on a roster. Although recruiting at times can be overvalued, it’s no coincidence that the last five national champions have had a four-year average recruiting rank in the top six nationally. The difference between the 16th- and 20th-ranked classes is slim, but there is value in knowing the difference between the 16th- and 50th-ranked classes.
Coaching tenure –- the final piece of the FPI puzzle -- allows FPI to capture the presence of a new head coach. It makes sense that a new head coach would decrease the correlation between the previous season’s performance and the future one (relative to returning the same coach). Depending on a team’s performance in prior years, a new head coach can have a positive or a negative impact on the projections.
What do the ratings mean?
Combining all of the factors above produces a predicted value on offense, defense and special teams, which represents the number of points that each unit is expected to contribute to the team’s scoring margin if it were to face an average FBS team on a neutral field.
LSU sits atop preseason FPI, but that does not mean the Tigers will win the most games in the FBS or make the College Football Playoff. The Tigers play a difficult schedule with road games against Florida, Arkansas and Auburn, so when the full season projections are released this spring, LSU might not even be favored to win the SEC.
Nevertheless, preseason FPI provides a vehicle to produce unbiased ratings that account for many of the same factors that voters tend to value when producing their preseason rankings –- past performance, returners, talent, etc.
Preseason FPI also is important from a practical perspective to allow the correct opponent adjustment in the early games of the season and helps with overall prediction accuracy going forward. Stay tuned for the next iteration this spring, preseason FPI 2.0, which will include full season projections and up-to-date starter information.