Predicting Wins – The Case for the Conferences

In my most recent piece, I took all the major conference teams and calculated predicted wins for the upcoming year based on the total strength of recruiting classes per Scout.com for the past for years. (see here).

I used a model based upon similar data for the 2006 season (see here). While the model was off by as much as 8 games in one case, it averaged less than a single game error for the 66 teams examined.

I have received much in the way of helpful comments and criticism. One of the most obvious flaws appears to be the predicted records of the Big East teams – it essentially has none of them doing better than .500.

The problem with the Big East is the same problem I initially had with the mid-majors – although their recruiting talent is not as formidable as the “bigger” conferences, since they play mostly themselves, someone must rise above and win. If you look at the prediction post, you will see that, although ranked 32nd overall, Pittsburgh has the best 4 year recruiting class in the Big East.

The problems as identified here is much the same as the discussion about the Big East last year and what it would take for one of its teams to reach the BCS title game. Essentially, it was agreed that such a team would have to be undefeated. As these teams lost a single game, they were ruled out. Thus, it was recognized that the Big East was “easier” than the other major conferences.

I have decided the best route for applying our model is likely on a conference by conference basis, as I did with the SEC in the last piece. To recall, the model predicted the following for the SEC –



I have now run a similar analysis for the Pac 10, the Big 10 and the Big East.

First, the Pac 10 –

What is notable about the Pac 10 is the distance USC has put between itself and its conference rivals in terms of accumulated talent, having nearly 60% more total points than California at number 2. However, the rest of the conference is remarkable close in talent, with the difference between Cal and the worst team (Washington State) being smaller than the difference between USC and Cal.

Thus, look for a USC sweep next year.

Next, the Big 10 –

Not much in it looks very surprising to me, although our Wisconsin fan lurking in the commentary may take offense. But Wisconsin loses a 5th year senior QB, as well as a number of seniors on offense and defense.

Finally, the aforementioned Big East –

I agree it seems unlikely to me that Louisville, with all their remaining talent and Brian Brohm at QB, is going to go 4-3 in the Big East. And although Brohm was a big time recruit in 2004 per Scout.com, he was easily the most talented in that class, and the class itself was not highly ranked. But, if we are to stay true to our model, we have to take what it says, no matter how unlikely it seems now (the Model would have had Louisville finishing 2nd in the Big East in 2006).

And once again, it is important to remember this is just a model, and one based purely on talent as defined by Scout.com at that. While talent is just one factor in deciding college football wins (with coaching and luck be two other obvious ones), I believe talent is the primary factor in wins over a season. Our model as applied to last year would have been highly predictive overall.

As applied to next year, we shall see.

Next - The ACC and Big 12