NFL Draft: Do BCS Champions Make NFL Busts?

by Henry Johnson

With the NFL draft less than one week away, teams are scrambling to determine which characteristics of potential picks matter and which ones donâ€™t. One possible trap to which teams might fall prey is an attraction to BCS champions. Players with national championship rings, after all, earn heightened media exposure and a reputation as winners, whether or not those features are predictive of NFL success.

To test whether college champions are overvalued, we can use data from Pro Football Referenceâ€™s draft page, which comes complete with a metric called Career Approximate Value. CAV is a measure of player performance that assigns the greatest weight to a playerâ€™s best year, then the second greatest weight to his second best year, and so on. It is derived from Approximate Value, another metric that Pro Football Reference created for player assessment.

In order to see the relationship between CAV and pick number (and in order to show how spectacular of a steal Tom Brady was), a scatterplot with data from 2000-2014 is shown below:

Thereâ€™s a clear negative relationship between pick number and Career Approximate Value, and once we transform the variables, the relationship can be estimated fairly well via linear regression. The transformation that properly takes care of the issue of heteroscedasticity turns out to be taking the square root of pick number and the square root of CAV.

To see whether winning the BCS National Championship increases a playerâ€™s chance of becoming an NFL bust, we can create a dummy variable that takes on a value of 1 if a player was fresh off a championship. Note that a player who, say, won a championship his junior year but not his senior year would register a 0 for the â€śChampâ€ť variable. All this variable tries to account for is the hot commodity label of having just won a title.

When we conduct linear regression for the square root of CAV using the square root of pick number and â€śChampâ€ť as our predictor variables, we get the following results:

While the coefficient on the championship dummy is negativeâ€”indicating that college champions are overvaluedâ€”the relationship is not significant, as the p-value of 0.683 shows. This is not to say that NCAA champions arenâ€™t primed for disappointing NFL careers, but the link isnâ€™t strong enough to draw conclusions with confidence.

At the end of the day, the data donâ€™t give a single, strong answer. The lack of clarity may be frustrating for teams considering using a late first-round pick on Ohio Stateâ€™s Devin Smith. Those franchises, instead, would be wise to use firmer findings of the past.

harvardsports

View all posts

• Mike says:

The Pick # vs. Approximate Value is an awesome chart. I’d love to know who some of the other outliers are, specifically:

1) Who are the players with approximate value right around 100 in picks 50 – 150?
2) Who are the two outliers at the end of the draft around AV 70ish?
3) Who are the picks roughly at 175 and 225 with negative approximate value?

Thanks for providing the great content!

• Henry says:

Hi Mike, thanks very much for reading!

1) The first is Lance Briggs, taken 68th with a CAV of 98, then Steve Smith, who went at 74 and had a CAV of 99. Those two are clustered together on the left. Then Jahri Evans was taken 108th with CAV 101. Jared Allen is the last one in that range, taken at 126 with CAV 96.
2) Marques Colston was taken at 252 and had CAV 69, and Scott Wells was taken at 251 with CAV 66.
3) Owen Pochman was taken at pick 216 and generated CAV of -2. And the CAV of -4 at pick 185 is… poor, unfortunate Ryan Lindley.

Thank you for the terrific questions!