Why does the FPI hate the Big Ten?

ESPN’s Sports Analytics Team released its preseason Football Power Index last week, and the Big Ten took a bit of whooping.

Only three conference teams landed in the Top 25 of the rankings. The highest was Michigan at No. 11 -- significantly further down the food chain than most early preseason rankings expect the Wolverines to be.

Poor rankings led to poor schedule strength. The five weakest schedules among Power 5 schools next season belong to Big Ten teams, according to the FPI. Only four Big Ten schedules are ranked in the Top 50 (none higher than Ohio State at No. 31) despite an East Division with three playoff contenders and some marquee non-conference games like Wisconsin vs. LSU and Ohio State vs. Oklahoma.

It’s important to remember the FPI is a statistical ranking determined by numbers and not people, but it does beg the question:

Why do the numbers hate the Big Ten?

The preseason rankings are built around four major components, each given different weight: performance in the past four years (with more emphasis on the most recent season), returning starters, recruiting rankings and coaching tenure. Most folks assume the recruiting rankings sink the Big Ten, but the first two factors actually have a more detrimental impact this season.

For example, Ohio State checks in at No. 17 on the list despite its 50-4 record over the past four seasons. The Buckeyes are three spots behind Notre Dame, a team they beat by two touchdowns in their most recent game. So why did they drop so far? Only six starters return from last season's team, which is the lowest number of any FBS team in the country.

"That’s a big thing impacting Ohio State. Michigan State is returning only 10, which is a pretty low number," said Sharon Katz from ESPN’s Stats & Information Group. "If they were returning an average number (about 13), I think they’d be way higher."

The returning-starter metric gives extra weight to teams that return a starting quarterback -- another factor that hurt the Big Ten this year. Michigan State, Michigan, Penn State and Wisconsin all suffered because they are searching for a new starter under center this spring.

It’s also important to note the "past performances" category isn’t based on a team’s win-loss record, but its efficiency on offense and defense. That’s how Northwestern (10-3 in 2015) lands at No. 52, one spot behind Boston College, which didn’t beat an ACC team last season. The Wildcats won their close games. The Eagles didn’t. But the teams had comparable efficiency rankings (terrific on defense, dreadful on offense). The thought is both will move toward the mean.

But, if the numbers tell us a 10-win team with plenty of returning starters is comparable to a three-win Boston College that lost the leader of its top-notch defense ...

Why is the FPI using those numbers?

Well, mostly because they work.

The main purpose of the FPI is to predict the winner of individual games. The rankings are only data points in a formula that determines who is more likely to win each matchup on a team’s schedule. And in that regard, the FPI has been pretty successful. Last season, it got 77.7 percent of its predictions correct (Oddsmakers in Las Vegas had a 78 percent success rate). The Prediction Tracker, a website that ranks rankings, had the FPI as the 12th most accurate system among the 68 that it tracks.

"If you’re a fan I completely understand, ... it’s all about the rankings," Katz said. "To us the rankings are a means to an end."

Can’t the FPI do better at ranking teams anyway?

Yes, of course. Predicting the actions of 20-year olds and the bounces of a prolate spheroid is difficult. The analytics team swings and misses on a few teams each season and tries to tweak its formula to account for those issues. For instance, last season they started given more consideration to programs that have experienced transfer quarterbacks replacing an outgoing starter at that spot.

There are problems that continue to vex those making the formula. How do you account for coaches that seem to always find a way to win? Teams like Mark Dantonio’s Michigan State and Pat Fitzgerald’s Northwestern outperformed their efficiency rankings, while Boston College clearly did not.

"We toyed with this offseason, how can we capture the 'Dantonio factor?'" Katz said. "I have a feeling that’s something in the next years we’ll try to dive into more. (Coaches like him) obviously have an impact as well."

For now, Big Ten fans will have to find solace in the idea that a few of the conference’s teams have a habit of outperforming the numbers. As for those weak strength of schedule rankings, though, they have only themselves to blame. Newcomers Maryland and Rutgers ranked 74th and 75th in the FPI this offseason. Only one Power 5 program (Kansas) was lower.