by Bill Connelly

Every few weeks this season, it is a good idea to check in on how the S&P+ rankings are coming together. Clearly it is early enough in the season that first, as with FEI, preseason projections are still playing a decent-sized role, and second, one good or bad game could alter a team's rating pretty significantly. That said, we're going to dive in nonetheless.

S&P+

As you may recall from previous editions of Varsity Numbers, the S&P+ rating is made up three components:

• Success Rates
• EqPts Per Play, or PPP
• The "+" adjustment (best described here)

S&P+ is designed to measure efficiency (Success Rates) and explosiveness (PPP), adjusted for strength of schedule ("+"). As you would expect three games in, the limited schedule is going to contribute significantly to a team's rating, but that is one of the reasons preseason projections are still being used for 50% of the ranking at this point. Projections were determined using similar factors as those used for FEI projections (five-year history, returning starters, etc.), along with some of the turnover figures discussed here. With more years of data (by the offseason, there should be complete play-by-play for at least 2005-09, and maybe 2004 as well), more factors can be taken into account, hopefully creating a very useful, accurate system of projections.

In theory, if the projections are any good, they level out some of the extremes in the three-week raw numbers. But with such a small sample size, these early rankings should be used more for simply figuring out who may be overlooked or underrated (in terms of national perception) instead of truly ranking all 120 teams in terms of performance.

So without further adieu, here is your Top 120 FBS teams. Listed below are the following measures:

• S&P+ and Rank
• Offensive S&P+ and Rank
• Defensive S&P+ and Rank
• Raw S&P margin (Offensive S&P minus Defensive S&P, unadjusted for strength of schedule)

These numbers do not take into account South Carolina's win over Mississippi Thursday.

Surprises and Thoughts

• Penn State has so effectively blown out the cupcakes on their schedule that their ranking increased despite strength of schedule.
• Seven of the Top 10 are from either the SEC or Big 12. No real surprise there.
• Thanks to projections and the team's decent performance against third-ranked Ohio State, USC has not fallen far despite the loss to Washington. Future performance and schedule adjustments will even that later.
• Clearly Oregon is still benefiting from its preseason projections. The Ducks only have a couple of weeks to start achieving before their ranking falls precipitously.
• Best five teams with losing records: Tennessee (14), Virginia (39), Troy (40), Michigan State (46), Nevada (51). No explanation whatsoever for Virginia being ranked 39th.
• Five lowest-ranked undefeated teams: Indiana (92), Colorado State (89), Arizona State (82), Texas A&M (75), Kentucky (69). Clearly if they keep winning, they will each show some upward mobility.

At the very least, this exercise should give you insight for some of the S&P+ picks made during the weekly Seventh Day Adventures article.

F/+

In this summer's Football Outsiders Almanac, there was one more figure shared with FEI and S&P+. It was the combination of the two, known in the book as F/+. For 2008, both FEI and S&P+ had correlations of about 0.83-0.85 to winning percentage. The correlation of F/+ was around 0.88. This makes sense, of course. Both measures have their advantages and disadvantages, and a combination of the two would cancel out some of the outliers.

The F/+ measure has been good early on at balancing out the S&P+ infatuation with Penn State, Georgia and TCU, and FEI's love of all things Auburn and West Virginia.

The goal is not simply to use Varsity Numbers as a space for posting weekly S&P+ (or F/+) rankings, but every few weeks (when the column ideas are running a little low ... ahem), checking in on these measures is a good thing. And we're working on weekly updates of both S&P+ and FEI in the FO statistics section

Posted by: Bill Connelly on 25 Sep 2009

13 comments, Last at 28 Sep 2009, 12:57pm by DaninPhilly