Comparing the final Football Outsiders F/+ projections to the AP poll and conventional wisdom.
28 Aug 2007
by Brian Fremeau
The Fremeau Efficiency Index (FEI) was introduced to Football Outsiders last season, and the details of its principles and methodology can be found here. Like DVOA, FEI rewards playing well against good teams, win or lose, and punishes losing to poor teams more harshly than it rewards defeating poor teams. Unlike DVOA, it is drive-based, not play-by-play based, and it is specifically engineered to measure the college game.
This 2007 college football season preview will not include an analysis of the percentage of offense your favorite team lost to last year's NFL draft. It will not include a discussion of coaching changes or hot seats. It will not appear on your local newsstand featuring a regional Heisman hopeful. And it will not cast a single vote in a preseason poll.
The Fremeau Efficiency Index (FEI), in fact, was never designed to preview college football before the bands take the field. According to its principles, the best teams in college football can best be determined by collecting and weighting the Game Efficiency results of games played in a given season. Such a retrodictive rating may very well help predict the outcome of late-season games and bowl games, but given the cycle of recruiting classes and coaching changes that seemingly shift the balance of power in college football every year, how reliable could FEI be in forecasting next-year results?
As it turns out, one of the most significant differences between the NFL and major college football is precisely the nature of their respective balances of power. The NFL deliberately fosters parity through the salary cap, scheduling, and the draft. In college, the elite of the BCS conferences fiercely feed their own power through recruiting prowess, lucrative television contracts, and control over scheduling. College rosters may include new starters at virtually every position every two years, but the correlation of year-to-year wins in college football (.59) is far stronger than that in the NFL (.24). Even stronger in college football are the year-to-year correlations of Pythagorean wins (.67) and FEI (.76).
Given the consistency of FEI ratings year to year, how much faith can be placed in previous-year FEI data to forecast next-year game outcomes in college football? Calculating the percent of next-year games won by teams with superior previous-year ratings (Forecasted Winning Percentage) produces the results found in Table 1. As a reference point, the accuracy of the weekly wagering line in picking game winners over the same period is also provided.
|Table 1: Forecasted Winning Percentage, 2004-2006|
|Season||Weeks 1-7||Weeks 8-14|
Forget depth charts, returning starters, coaching vacancies, off-field incidents, injury updates, new coordinators, and redshirts for now. There won't be many questions as to who will populate the top of the national scene until the season's second half. The games themselves, however, will be noticeably impacted from the outset by two significant rule changes made this past off-season.
The first change retracts last year's controversial rule dealing with the game clock and changes of possession, one that aided clock-killing situations for teams in the lead, left many coaches scratching their heads in critical situations and spurred fan angst all season long. The biggest knock against the 2006 rule was the apparent loss of football. Indeed, the average game lost approximately 14 plays and 1.5 possessions last season. The 2007 rule change should restore those plays and possessions, but will it actually improve the games themselves? The average margin of victory in games was reduced by almost four percent from 2005 to 2006, and the total percentage of games decided by 16 points (two possessions) or fewer increased by three percent last year. The 2006 rule may have caused more confusion for coaching staffs, but it also seems to have created more competitive contests. College football fans may get more football in 2007, but more of it may be garbage-time football.
The most significant new rule change moves the kickoff in college football from the 35-yard line to the 30. Undoubtedly, the number of kickoffs downed in the end zone will drop dramatically this season, likely resulting in more exciting returns and better average starting field position for offenses. Scoring across the country will likely get a boost from this rule, and not exclusively from the receiving teams. More kickoff returns will contribute to more turnovers on kickoff returns, too, giving opponents more opportunities for short field possessions going the other way. The average starting field position for offensive drives in 2006 was the 30-yard line, and Figure 1 illustrates the success rate of scoring in college football from every starting position on the field last season.
|Figure 1: Offensive Efficiency from Field Position (2006)|
|Efficiency = ( Points / 7 ) / Possessions|
I will be keeping track of the effects of the new kickoff position on my site throughout the season. Fans in the stadium may be swept up in the excitement of the returns themselves, but should pay even more attention to the resulting field position. Game Efficiency and FEI are built on individual drive success, and the new kickoff rule may dramatically impact those results in 2007.
As mentioned above, FEI is not interested in casting a vote in a preseason poll. The judgment of voters in late-season polls is often too heavily influenced by preseason poll position, and the tendency of voters to anchor teams to positions throughout the season aggravates the problem. That said, like FEI forecasts, the AP and USA Today preseason polls tend to reinforce the power balance in college football at the top rather than challenge it, and thus project pretty much the same teams for elite success in 2007. To produce the FEI forecasted records for 2007 listed below, I combined FEI data from the past two seasons to project the game outcomes for the first half of the season. This projection method produced a 72.8 percent record in selecting winners over Weeks 1-7 in 2005 and 2006.
The top eleven teams in the country according to the AP and USA Today polls are expected to perform well over the first half of the season, losing only games scheduled against other top-eleven teams.
(Alphabetically, with projected record on Oct. 14):
Ohio State (7-0)
Virginia Tech (5-1)
West Virginia (6-0)
Outside of the top eleven, however, FEI disagrees with the preseason polls about the likely fortunes of several ranked or unranked teams:
Auburn (No. 18 AP, No. 14 USA Today)
FEI Forecasted Record on Oct. 14: (6-1)
BYU (Unranked AP, Unranked USA Today)
FEI Forecasted Record on Oct. 14: (6-0)
Notre Dame (Unranked AP, Unranked USA Today)
FEI Forecasted Record on Oct. 14: (5-2)
UCLA (No. 14 AP, No. 17 USA Today)
FEI Forecasted Record on Oct. 14: (3-3)
Penn State (No. 17 AP, No. 18 USA Today)
FEI Forecasted Record on Oct. 14: (4-3)
Florida State (No. 19 AP, No. 21 USA Today)
FEI Forecasted Record on Oct. 14: (3-3)
The first published FEI Ratings for 2007 are scheduled to appear here on Football Outsiders in mid-October, at which time it will be open season on these FEI forecast hits and misses. Until then, feel free to drop me a line here at FO, or visit my site for drive efficiency data throughout the first half of the season.
84 comments, Last at 16 Oct 2007, 2:42pm by Bob Day