clock menu more-arrow no yes

Filed under:

What predicts recruiting success? Long-term success, of course.

Comparing S&P+ and previous recruiting momentum to predict future recruiting success

Jason Getz-USA TODAY Sports

Last week I took a look at the concentration of recruiting power in college football and found that there is a very small and consistent group of elite recruiting schools. As others have established, these elite recruiting schools often produce championship contenders.

I wanted to run a few simple statistical tests to further examine the concentration of recruiting power. Using 247 Composite team ratings (compiled by Brian Fremeau) and the S&P+ rankings, I looked at what variables best predict recruiting power.

Recruiting momentum is strong

Often one of the best predictors of success in college football is how the team did the previous year, and it's the same case for recruiting. I ran a simple regression between 247 Composite recruiting rankings between 2013-2015 and the 2016 recruiting rankings and found that there seems to be a positive, statistically-significant relationship between the 2013-2015 average recruiting rankings and 2016 rankings.

t Stat P-value R-sq
2.5 .013 .87

And going by the R-squared of .87, the three-year recruiting average explains 87% of the variation in 2016 recruiting class rankings. This fits perfectly with the result we had last week where only 16 total teams have had a top-10 recruiting finish over the course of the last four years.

Last year's on-field performance matters a lot...

I also took a look at whether on-field performance matters for next year's recruiting. My thought here would be that playoff teams and other successful teams from the 2015 season could use their strong finish to wrap up elite recruiting classes. Obviously that was the case with Alabama, but teams like Clemson and Ohio State also sat on top of the recruiting rankings after being some of the best teams in 2015. So I looked at whether 2015 S&P+ rankings affected 2016 recruiting rankings and found that again there is a positive, statistically significant relationship between the two variables.

t Stat P-value R-sq
41.7 .00...1 .49

From the results above, we can see that there is strong evidence that the positive relationship between 2015 S&P+ and 2016 recruiting rankings is not random, but it explains just under half of the variation in the recruiting ranking results. So as a predictor for recruiting success, it seems to be less predictive than previous recruiting success, but it is almost definitely less likely to be a random relationship.

...but consistently elite teams do even better at recruiting

This shouldn't be surprising. I also tested whether a five-year S&P+ average is more predictive than last year's S&P+ average and found that consistently quality programs (over a five-year period) predict recruiting success even better than last year's.

t Stat P-value R-sq
49.1 .00...1 .63

So this relationship is even less likely to be random but explains a larger percentage of the variation in the 2016 recruiting rankings (though, still less than previous years' recruiting did).

These results suggest that elite recruits might be drawn to schools that performed well the previous season but are more likely to put their faith in schools that have a consistent pattern of success. Five years is a long time in college football. Coaches can come and go (Urban Meyer was in and out of Florida after five seasons), teams can go from playing in the national championship to firing their coach and going 6-7 (Texas in 2009 to firing Mack Brown and Charlie Strong's first year in 2014). So to be consistently solid on the field over that span is impressive and a good indicator to recruits that they are likely to have similar success by choosing that school.

It also suggests that there's still a break between recruiting and on-field performance, since the R-squared values were still much different between the first and third tests. So I did a final multiple regression comparing the 2013-15 recruiting averages with the five-year S&P+ average to see if either variable dropped out when taking the other into account.

t Stat P-value R-sq
S&P+ Avg 2.46 .015 .88
Recruiting Avg 15.9 .000 .88

So both variables are significant in explaining the variation in 2016 recruiting results, though adding the S&P+ average only improves the model by 1%. So in the end, recruiting success seems to be the best predictor of recruiting success. Obviously this only looked at the 2016 class, so more data is needed, but it does seem to validate some conventional wisdom about recruiting powerhouses.