by Brian Fremeau

It’s been a long offseason, and I think we’re all ready for some football. The last few months have been peppered with salacious scandals and conference crises, but there are finally games to be played and game data to be consumed. I’ve taken the opportunity in the offseason to thoroughly examine the ways in which I process college football drive data, and I’m particularly excited to get a new season underway. There are new and enhanced metrics I’ll be debuting in 2011, and I’m ready to get started.

But before we get into the new stuff, there are some things that haven’t changed at all. The Fremeau Efficiency Index (FEI) college football ratings will continue to be published weekly here at Football Outsiders and the methodology to produce the overall FEI team ratings has not been modified. Game Efficiency still represents my summary measure of each team’s offense, defense, and special teams effectiveness over the non-garbage drives of a game. Game Efficiency continues to be the raw metric at the core of the FEI ratings, and my opponent adjustment methodology has not been altered this year. I’ve been pleased with the consistency and quality of the overall FEI ratings as well as the game projections and complementary metrics (strength of schedule, mean wins) that are derived from those ratings.

What has changed is the way I have carved up drive data to produce separate offense, defense and special teams efficiency metrics. The changes aren’t so significant that they dramatically shake up Offensive FEI or Defensive FEI, and the principles behind the metrics haven’t changed at all. But there were gaps last season -– bits of drive or field position value that was unaccounted for in the splits data. I’ve closed those gaps and can now distribute that fractional drive value to the offense, defense, or special teams unit that was responsible for producing it. Let’s review the principles and process.

Every possession has automatic expected value. A typical college football game consists of 24 non-garbage game possessions (12 for each team) and each team is expected to score about 23 points. In other words, every drive is worth a little less than two points, automatically. Doing nothing other than possessing the ball is, on average, worth something.

We can be even more precise. Following the opening kickoff of each half, the automatic expected value of the ensuing possession is 1.6 points. On every other non-overtime possession in a game, the expected value is 1.9 points. What accounts for the difference? Starting field position is better, on average, on non-opening-of-half drives. Which leads us to the next principle.

Starting field position value is earned. Drives that start in opponent territory versus drives that start on a team’s own side of the field can have very different scoring expectations. When a team starts a drive at midfield, it should expect to score 2.4 points -– more than the automatic expected value we had just described. And that extra value was earned by a combination of the offense, defense and special teams play during the preceding possessions. Likewise, the automatic value lost due to starting a drive deep in a team’s own territory is also shared by those units.

The value distribution for each unit in each situation varies based on the specific factors in play. Kickoff returns are measured against average resulting field position. Punt return expectations are a function of the return unit and the drive-ending field position of the previous possession. And who is responsible for the drive-ending field position?

The result of a drive impacts subsequent possessions. This is the gap that I needed to address this offseason. Previously, I measured offensive success on a drive in isolation; results versus expectations of every drive were calculated in a vacuum. But a fraction of the scoring expectation of the opponent’s next drive (and sometimes, the team’s own drive after that) is at stake as well. By closing the gap, the fractional next-possession value at stake is now attributed to the offense, defense and special teams units that earned it. This was covered in an FEI column two weeks ago.

The offense, defense, and special teams efficiency splits I produce are based on these principles and the measurable average expectations of every possession and possession change event.

So how does this approach change the offensive and defensive FEI ratings from last season? Actually, over the course of the season the changes aren’t particularly dramatic. At the game level though, the differences are a bit more apparent. For instance, in the Rose Bowl last season, TCU’s offense scored three touchdowns on eight total possessions against Wisconsin. But part of the key to TCU’s victory was its consistent success on non-scoring drives. The Horned Frogs went three-and-out only once that day, moving the chains and forcing the Badgers into poor field position throughout the second half. The incremental value of each of those non-scoring drives improved TCU’s offensive efficiency for the game by by about 10 percent in our new methodology. Likewise, Wisconsin's defensive efficiency in the game drops, having been unable to get TCU off the field throughout the game.

On the other end of the spectrum, teams that created field position advantages through defensive and special teams play are given an extra boost. The USC defense set up frequent scoring opportunities for the offense against Notre Dame late last year, positioning four drives following turnovers in Irish territory. Though USC's offense sputtered, it's defense earned extra value on the possession changes that never materialized on the scoreboard.

As mentioned, the impact on season-long OFEI and DFEI rankings were not significant for most teams. Teams that were strong in isolated possession success were generally just as strong when accounting for the fractional field position value now added in. The following tables represent the new rankings for 2010 OFEI and DFEI, with last year's ranking provided for comparison.

The other significant addition to this year's FEI data will be expanded special teams ratings. I broke down punt, kickoff, and field goal efficiencies last season, but I've been meaning to combine special teams into a single metric. We'll dig into those numbers in more detail in the coming weeks, but for now, I've simply added up the cumulative value created by each team's field goal units and ranked every team accordingly. As may have been expected based on field position numbers I've shared recently at my blog (here and here), the LSU Tigers created more special teams value than any other team in college football last year.

All of which sets the stage for 2011. Readers that purchased Football Outsiders Almanac 2011 know that the F/+ projections favor Alabama as the favorite to win this year's national championship. This year's FEI projections, based primarily on Program FEI, returning starters and recruiting data, also love the Crimson Tide. As has been the case in the past, the projections will play a part (reduced each week) in the FEI ratings until week seven. FEI forecasts for every FBS game will also be posted on my site.

Posted by: Brian Fremeau on 01 Sep 2011

10 comments, Last at 11 Oct 2011, 2:12pm by