For the past three seasons (more or less), I’ve attempted to publish a weekly forum piece diving into advanced analytics for college football. This started as a curiosity exercise to determine how far from reality the AP pollsters can be in their voting.
Enter Bill Connelly and Brian Fremeau, who started the site Football Outsiders. These very smart gentlemen log every snap of every FBS college football game, and run all the statistical data through some rather large gonculators and spit out highly predictive quality ratings of the teams. Connelly produces the S&P+ and Fremeau produces the FEI. They combine these ratings into a single, comprehensive rating that they call "F/+". If you’d like to know more about their systems, I encourage you to spend some time poking around their site. Their ratings appear to be one of the primary sources for initial Vegas betting lines, so they're pretty good.
If you don’t like advanced stats, advanced analytics or you’re unwilling to believe that Team A, which lost head-to-head to Team B or has more losses than Team B, can actually be a better overall team than Team B, then this probably isn’t your thing. I like it, because these guys look at raw data, every snap, down-and-distance, etc., and ignore garbage time and clock kills at the end of halves. I also really appreciate them because of this:
A team is rewarded for playing well against good teams, win or lose, and is punished more severely for playing poorly against bad teams than it is rewarded for playing well against bad teams.
The first three to six weeks of F/+ ratings are only slightly better than the AP Poll. They use a lot of historical data to develop the initial ratings (about 50% returning offensive and defensive production, 25% two-year recruiting average, and 20% recent history – I know that doesn’t add to 100), and they try to avoid big shifts in the ratings by only adding new data a week at a time and slowly removing the weight of the preseason data. By week six or seven, the preseason bias disappears, and it’s all in-season stats after that.
Let’s get analytical (Ramzy, please don’t sue me.)
Preseason Top 50
The only preseason rankings/ratings we have to compare are S&P+ and the AP Poll. Next week I'll add FEI and F/+, and when CFP rankings come out they'll be listed as well.
Week Zero Takeaways
Top 10: The AP Poll has a few anomalies in the Top 10, as compared to S&P+. The most obvious is the Buckeyes sitting at #1 in the more objective measure, i.e. S&P+. The other glaring difference is Wisconsin, which AP voters love at #4 but S&P+ has all the way down at #11. The rest of the Top 10 seems reasonably close for preseason.
Overrated: AP Pollsters ranked the following teams at least 5 or more places above their S&P+ ratings: Stanford (+7), TCU (+6), Texas (+7), Florida (+13), Va Tech (+14), Iowa State (+9) , Boston College (+7) and West Virginia (+31). The most egregious of these is WVU, which the human poll ranked at #17 while S&P+ put the ‘Neers all the way down at #48. Basically, the pollsters love them some ACC, sort of. The strange part is the AP Poll completely left out Louisville, Wake Forrest and Duke, which are all projected higher than BC and WVU by S&P+.
Underrated: You may not believe this, but the AP Poll actually underrated seven SEC teams as compared to their S&P+ preseson ratings. (Maybe the love affair is dying.) S&P+ has LSU 10 places higher than the AP Poll, Miss State four spots higher, Texas A&M 10 places higher, and Auburn three spots higher. S&P+ also has Ole Miss, Mizzou and Arkansas in its Top 40, but none of them received a single vote in the AP Poll.
Notable metrics: A couple things to look at in S&P+, if you’re trying to compare teams to each other. The simplest comparison is to take the difference between their overall S&P+ projected. For example, OSU vs. ‘Bama shows about a three-point spread.
If you want to dig a little deeper, compare the Off S&P+ of one team to the Def S&P+ of their opponent. This is where you can get a better idea of how they match up. What become very apparent are teams that are unbalanced. Oklahoma is the perfect example: ranked #1 in S&P+ offense but #83 in defense. Historically, that's a bad formula for making it to the CFP.
For the four years of data we have on the CFP, the teams that make it in are generally Top10 in both S&P+ offense and defense by the end of November. Using the preseason S&P+, the four most ‘balanced’ teams are OSU, ‘Bama, Washington and UGA. Auburn, PSU and Clemson would be the next tier.
AP “Top 25” Match-ups
#9 Auburn vs. #6 Washington – This is a great match-up on paper. UW should be a 2-point favorite, but they play in Atlanta. I think the 3:30 start favors the west coast team, so I’m going with the Huskies in a tight one.
#14 TTUN @ #12 Notre Dame – Another spectacular match-up according to S&P+, with less than 1-point separating these teams. The Irish have weird voodoo at home, and I hate their opponent, so I’m picking ND to “win” by a missed PAT.
#20 Va Tech @ #19 FSU – Did I mention that the AP poll ranked the Hokies 14 places higher than their S&P+ rating has them? The Noles are a 5-point favorite according to S&P+, and they should win this one at home. I hope they're prepared for a "Bear" front.
As always, I welcome your comments.
(Oh, and Wildcats by a FG tonight, if you buy into S&P+.)