Now that the regular season is complete, I took a look at how each team fared relative to the talent they have on hand. Taking the composite rating of each SEC team’s roster talent, as calculated by the Rival’s 247 Composite website, I analyzed how each team’s overall win percentage aligned with their talent. Obviously, we would expect teams with more talent to win more than those with less talent. But by isolating the talent as a variable, and then viewing the results of the season, we can see how good of a job each coach has done. To be sure, there is an element of randomness (injuries, etc.). However, this analysis gives you an idea as to how a team performed relative to its talent level.
In the chart above, you can see the regression line through the middle of the chart. This represents the expected value of win percentage (along the vertical, or “Y” axis) is it relates to the 247 Composite Talent average per team (horizontal, or “X” axis). Teams below the line had a lower winning percentage than their roster talent would predict. Teams above the line performed better than their roster talent would predict. And, of course, those teams with scores on or at the line performed as expected. The distance from the line indicates the degree in which a team under-performed or over-performed.
The table below shows the data. Kudos to all the teams that exceeded the expectations. Detention for those that didn’t, especially Tennessee and Auburn.
For our stats friends, the model is significant and has an adjusted R² of .333, which means talent rating accounted for 33.3% of the variance in win percentage, with 66.7% of the results coming from factors other than talent rating. If you care to play around with the numbers (hey, SLR is fun because it is easy), the β = -4.009, ‘Talent’ coefficient (unstandardized) = 0.053. In this model, if you’re Florida for example, with a roster talent of 88.28, you would multiply 88.28 with 0.053 (giving you 4.68) and subtract 4.009 from that, giving you .670. So, you’re expected win percentage with be 67%, which is 8 games. So, Florida out-performed their talent by one game this year. Mullen is good at coaching sports. The table displaying expected win percentage based on talent is below:
Like the Florida example above, you can see how it all plays out for each team in this table. Alabama won one game beyond expectations (so they are above the line in the scatter plot above), while Arkansas lost 4 more games than their roster talent would predict. Of course, a ‘zero’ means they performed at expected levels. Great job this year by Kentucky, winning 3 games over their talent level.