Strength of schedule thoughts

Recently, I’ve seen a couple of posts attacking Jeff Sagarin’s strength of schedule (SOS) computer model, one at Saurian Sagacity (another excellent Gator blog, by the way) and one at AOL’s ACC Fanhouse. Both score some telling points – particularly with regard to the fact that Sagarin won’t disclose the methodology behind his rating system.

But I do want to take issue with one item both bloggers raise as an indictment of Sagarin, namely, that he ranks the PAC-10 as the top conference in his SOS rankings. It’s a criticism I see levelled regularly on the Internet and I’m not sure it’s a justified one. In fact, I can think of a good reason why it may be entirely valid.

First, though, let me say that there’s a difference between saying that a particular conference’s schools play the toughest strength of schedule and saying that a particular conference is the toughest conference to play in. The former simply takes into account which opponents appear on the schedules of the conference members. The latter would be more focused on the strengths of the top teams in the conference, what the conference venues are like to play in, how bad the weakest teams in the conference are, how the teams in the conference fared against top non-conference opposition, etc.

Given that, I don’t have any problem stating that I believe the SEC was the best conference in college football this season while also believing that the PAC-10 had the best SOS of any college football conference in 2006.

The reason that I alluded to earlier as to why Sagarin may not be off base here is a decision the PAC-10 made when the regular season was expanded to twelve games. That conference decided to add a ninth conference game to every member school’s schedule. That’s at least one more conference game than is played in any other major conference right now (the Big East teams play only seven conference games).

With regard to how this plays in to Sagarin’s computations, I would think that one significant factor would be addition by subtraction – the PAC-10 filled the twelfth game void with another conference game, while some schools filled that game with a much lesser opponent.

In fact, looking over the PAC-10 schedules for 2006, I could only find three 1-AA schools listed as opponents for the entire conference. I’m not saying that PAC-10 schools didn’t schedule any weak sisters (of course they did), just that, because of the ninth conference game, they scheduled fewer of them. You might take note that Alabama played two winless non-conference teams this year, or that Purdue’s first two non-conference opponents went 3-20 this year (and Purdue lost to one of them!) or that none of Wisconsin’s non-conference opponents had a winning record to appreciate my argument.

And it’s not like Sagarin is alone in his ranking the PAC-10 first. Massey and Howell, for example, do as well.

I expect that any major conference that joined the PAC-10 in going to a nine conference game schedule – and I’d be thrilled to see Georgia drop the likes of Western Carolina for another SEC West opponent, as an example – would see their SOS rating climb.


Filed under College Football

7 responses to “Strength of schedule thoughts

  1. Hi, Thanks for the link.

    The thing with Sagarin is that not only does he say that the Pac 10 is the strongest conference, he has actually assigned the top 10 strength of schedule scores to the 10 Pac 10 teams.

    That means that, according to him, there isn’t a team in the country that has played a tougher schedule than ANY (I repeat ANY) of the Pac 10 schools.

    Your argument about 9 conference games may explain why the conference overall is highly rated, though I doubt it, just because an opponent is a conference opponent doesn’t mean that it’s a good opponent. Florida for example played conference foe Vanderbilt that isn’t as good as non-conference opponent Southern Miss, a bowl team. The Pac 10 has it’s share of bottom dwellers too.

    But your argument doesn’t account for how Stanford’s (or any of the other Pac 10 schools) schedule is harder than for example the Florida Gators’ schedule. Remember Florida played 9 conference games too. In fact of the 12 teams in the SEC, Florida played the other 9 top teams. They didn’t get lucky and draw Ole Miss or Miss. State this year.

    It strains credulity to believe that top to bottom every team in the Pac 10 had a stronger schedule than ANY other Div. 1A team.


  2. I was just thinking about this some more and you may be onto something about the number of conference games that the PAC 10 teams played but not the way you were thinking about it. Pac 10 teams played 75% of their schedule against conference opponents. As a conference they played 10 less out of conference games than they would have if they only played 8 conference games (like most other teams).

    Since conference games tell us nothing about the strength of the conference (all the teams could be really good or they could be really bad or something in between) out of conference games become very important. With a smaller sample size of out of conference games the Strength of Schedule beomes less stable and dependent on fewer outcomes.

    I think that USC’s manhandling of Arkansas early in the season and their beating of overrated Notre Dame is causing all of the Pac 10’s SOS numbers to be skewed higher. Saagarin, like most of the computer polls uses not only the W-L % of a team’s opponents but of the opponents opponents.

    With the so far pathetic showing the PAC 10 has had in bowl season it’s possible that the SOS rankings could be changed drastically.

    I’ll be watching.


  3. Thanks for the responses.

    I wasn’t really attempting to defend Sagarin in his entirety, so when you say “It strains credulity to believe that top to bottom every team in the Pac 10 had a stronger schedule than ANY other Div. 1A team”, I’m not going to argue with you at all. But that’s not the same thing as saying that the conference as a whole has had the best SOS in 2006.

    I do think you and I agree on the primary impact that going to a nine game conference schedule has on the SOS calculations (although you say it in better detail than I did).

    But here’s something else to chew on: how much effect on the math does the size of the conference have? What I mean by that is that the PAC-10 is still a ten team conference. To me, that means that (1) there’s a better chance in a given year that they’ll have fewer truly wretched conference teams that would drag down the SOS numbers; and (2) the PAC-10 as a whole has fewer scheduling slots to fill with non-conference opponents (thirty) than do other major conferences (the SEC, for example, would have forty-eight). The more of these slots, the greater the odds of crappy opponents hammering a conference’s SOS, right?

    The latter point would tie in neatly with yours about USC’s schedule and the impact it may have had on the SOS numbers.

    Anyway you look at it, there has to be something going on here, because, as I noted in my post, several other computer models rank the PAC-10’s SOS #1. And I can’t think of any other good reason to justify that, other than the nine game conference schedule.


  4. Pingback: Get The Picture

  5. Well here’s the thing about Strength of schedule. You may play 5 really good teams and 7 weak sisters and have a weaker SOS score than a team that played 12 middling teams. It depends on the formulas used.

    Also when you get into averages it gets tricky. Florida played one more game than say Stanford. The game against the weakest opponent gets averaged in there bringing Florida’s SOS down. (I’m only using Florida as an example because it’s the case I’m most familiar with). But you can make an argument that when comparing teams with different numbers of games played you should put them on an even ground. In other words what were the 12 best teams Florida played and stack them up against the 12 best Stanford played.

    The problem with scheduling out of conference game as you know is that in big football programs you try to schedule as many home games as possible to generate revenues. Conference games require return games. So every year SEC teams (as an example) must play 4 home and 4 away. That only leaves 4 slots on the schedule. Throw in an out of conference rival (like FSU for the Gators and Tech for UGA) that will also require a home game every other year and that leaves you with 3 slots. Ideally, you want those 3 to be home games. Who is going to give you a home game without demanding one in return? The Washington Generals, that’s who. Weak sisters that come in for a payday.

    The SEC team rightly or wrongly feel they play the toughest schedules in the land. So they don’t worry about putting legit out of conference teams on their schedule. The problem is that OOC games are the only way the computers can judge the relative strengths of the conference. Think about it. The PAC 10 plays 45 conference games. That’s 45 winners and 45 losers. No way to judge how good the teams in the conference are relative to other teams.


  6. What’s implicit in your argument is that the PAC-10 is a crappy enough conference that its own lack of quality more than offsets the tougher OOC schedule it plays.

    While I don’t think that overall the PAC-10 is as good a conference as the SEC has been this year, I don’t think the gap between the two is as large as you think it is. Let’s face it – ultimately the difference between Southern Cal and Florida right now is a blocked South Carolina field goal. And I think the PAC-10 is the only conference that has two wins against OOC opponents appearing in BCS games.

    That doesn’t make them the best, but the PAC-10’s been a better conference than the ACC and the Big East this year. Probably better than the Big 10, too.

    And as for your point about SEC schools and scheduling legit OOC games, for every Arkansas (their ’07 schedule will probably destroy the SEC’s SOS ranking all by itself), there’s a Georgia scheduling an Oklahoma State, an Alabama scheduling a Florida State and a Tennessee scheduling a California. To me, that twelfth game is found money – some schools are going to spend it one way and some another.


  7. I think these Bowl Games (I think all of them are interconference) could do a lot to shake up the SOS scores at the end of the season.

    My only point is that with such a small sample size of out of conference play going into bowl season, the picture is kind of distorted. Also consider things that computers can’t take into account. USC beat Arkansas handily in week 1. The Hogs played the Gators tough in the SEC title game. Do you think the Gators played the same team that USC did? The computer says yes but McFadden didn’t play against USC. He’s Arky’s best player. Also the computers can’t take into account that a team can improve over the season (especially a young team). So Arky was a stronger team at the end (my contention) than at the beginning but on paper USC had a better win against a common opponent. If Arky were to lose to Wisconsin, they’d drop to 3 losses making USC’s win less impressive. When Notre Dame gets demolished by LSU (pray to God) likewise that win for USC will be less impressive. If USC were to lose to Michigan (a very distinct possibility) then UCLA’s and Oregon State’s and every other Pac 10 school’s SOS would take a hit.

    In other words all the precincts haven’t been counted yet. Let’s wait until the fat lady sings before we pronounce one conference better than another.

    Congrats to the Dawgs on the win tonight. Also to Kentucky for beating Clemson.