Wins & Recruiting

Last year I took a look at how South Carolina recruiting has stacked up against the rest of the SEC East and Clemson. I wanted to take it a step further by expanding the list of teams to include the entire SEC and ACC and look at how recruiting is affected by wins in previous years. This grouping basically includes all the major programs in the Southeast, and typically accounts for ~50% of the Top 25 in the recruiting rankings in a given year. The general belief in college football is that the rich get richer, and that recruiting acts as a positive feedback loop, i.e. more wins lead to better recruits which lead to even more wins. In reality recruiting is affected by a variety of factors. Some teams have more tradition, some schools are in regions with better demographics and access to top talent, and some teams are known to flat out cheat more than others. A splashy new coaching hire can also bring immediate benefits on the recruiting trail. All these other factors are hard to quantify, but win totals are readily available, so let's see what we can discern with those.

The rankings I used this time are the team composite rankings from 247 Sports, which are generally considered to be the most accurate as they pool data from multiple scouting services. I went back as far as I could (the year 2000), at which point the rankings are a bit wonky and max out at 44. For the analysis below I only used recruiting rankings from 2002 onward as those appear to be reliable from that point. Here's some other notes about the data:

  • Teams included are all SEC and ACC teams as those conferences are currently constituted. So all years for Texas A&M, Missouri, Miami, Virginia Tech, Boston College, Pittsburg, Syracuse, and Louisville are all included despite those teams being in other conferences for a portion of the years being reviewed. That also means no Maryland.
  • Years 2003 and 2004 only had 11 regular season games, as they were sandwiched between the 2-year trial of 12 games in 2001 and 2002 and college football officially moving to a 12-game regular season from 2006.
  • I looked up the RGB and Hex values for the set of official team colors for every team, and used those to customize the markers for the data points in the charts below. As you can imagine, this took quite a bit of effort and a not insignificant amount of time to accomplish. To save space and make each chart more legible I've included a single legend for each data point in Figure 1 below. Please refer to this when looking at the following charts.

Figure 1. Legend showing all teams included in the following charts and their associated marker using official team colors.

Preceding Year Wins and Recruiting

I believe the prevailing thinking amongst college football fans and pundits is that a winning (or losing) season leads to an immediate impact on the recruiting trail, so in Figure 2 below I've plotted the 247 Composite Team Rankings against each team's win total for the year immediately prior. Figure 3 shows the same data, but is more narrowly focused by showing only teams with Top 25 recruiting rankings.

Figure 2. 247 Composite Team Recruiting Rankings for each year over the period 2002-2018 vs each team's win total for the preceding year.

  • There's currently 65 'Power Five' teams in existence, so the worst you'd ever really expect a team from the SEC or ACC to finish is around 65th in the recruiting rankings. There are some notable exceptions (looking at you, Kentucky and Wake Forest), but you can certainly see a ceiling around 65 in the data.
  • As you'd expect, no team that has a 13+ wins in this group finished outside of the Top 20 in recruiting. On the other side, no team that had 2 wins or less finished inside the Top 20.
  • Apart from that, there's not really a clearly evident trend here. You can win between 3 and 11 games and end up anywhere in the Top 80 in recruiting.
  • The average number of wins in this data set is 7.36, with the mode being 7. The average recruiting ranking is 31.50, with a mode of 11.
  • Two easily identifiable outliers in the data are Louisville and Missouri. Both finished two seasons with 12 wins but ended up outside the Top 25 in recruiting. Both schools are located in areas not known for producing football talent, and Louisville's 12-win seasons came while they were a member of the Big East conference.

Figure 3. 247 Composite Team Recruiting Rankings (Top 25 only) for each year over the period 2002-2018 vs each team's win total for the preceding year.

  • There's not a lot more that can be determined from looking at Figure 3 other than just getting a closer look at who's in the Top 25. A team can win as few as 2 games and end up in the Top 25, although that is much more likely to happen with 8+ wins.
  • Alabama dominates the bottom-right of this chart. The bottom-left has a group of suspicious characters that will be discussed in further detail later.
  • Over the 17 year period referenced, only Georgia, Florida, LSU, and Florida State have finished in the Top 25 in recruiting every year. Auburn has done this 16 times while Tennessee, Alabama, and Miami have done it 15 times. South Carolina has 13 Top 25 finishes over this period, and Clemson has 12.
  • Not a surprise, but worth pointing out, is the fact that the SEC dominates the ACC in terms of Top 25 recruiting rankings. Of the 222 data points in Figure 3, 154, or 69.4%, belong to SEC teams.

Two Years Prior Wins and Recruiting

I've heard on multiple occasions that given the time required to build relationships with players and their family members, a big winning season has more impact on the recruiting class two years later than it does on the class immediately following the season (i.e. win totals for the 2000 season impact the 2002 recruiting class more than they do the 2001 class). In Figures 4 and 5 I've plotted recruiting rankings against wins from the season two years prior to see if this postulation is supported by the data.

Figure 4. 247 Composite Team Recruiting Rankings for each year over the period 2002-2018 vs each team's win total for the season two years prior.

  • At first glance Figure 4 looks very similar to Figure 2, but upon closer inspection the data does appear to be more closely grouped towards a central trend. In particular the 7 to 9 win range exhibits a denser grouping of data points.

Figure 5. 247 Composite Team Recruiting Rankings (Top 25 only) for each year over the period 2002-2018 vs each team's win total for the season two years prior.

  • As we saw with regards to Figure 4, there's not much to differentiate Figure 5 with Figure 3 above. Alabama has slightly less of a stranglehold on the bottom-right, while Ole Miss (2013) and Auburn (2014) look very suspicious by themselves in the far bottom-left. Both of those recruiting classes occurred one full season after a coaching change.

Overall I'd say there's maybe a little bit of truth to the thinking about wins having more impact two years down the road, but there's not anything definitive in this analysis, at least in this view of the data. Given the recent addition of an early signing period in December, I expect wins from two years prior to be much more relevant to recruiting going forward, as most teams now have a vast majority of their recruiting class locked in before bowl season.

Preceding 4-Year Win Average and Recruiting

In my personal experience, I didn't really become aware of the wider world of college football until around 9th grade. I certainly followed the Gamecocks in depth (as much as was possible pre-Internet at least) and generally knew what teams were historically good, but I really didn't know much about what went on beyond the borders of SC. I assume that I was a fairly typical young guy and that not much has changed in the past 20 or so years. I'm sure that as these potential recruits start playing more competitive football and getting contacted by coaches their awareness of the college football landscape increases dramatically, and when it's finally time for the some of them to choose a program their impressions of teams are based mainly on what has happened since they entered high school. Four years ago I would often see a recruit say something like 'South Carolina always beats Clemson' in an interview. Clemson's current run of success would likely result in a high school senior not realizing that Clemson is a historically mediocre program. In Figures 6 and 7 below I've plotted the recruiting ranking against the average win total for the previous 4-year span.

Figure 6. 247 Composite Team Recruiting Rankings for each year over the period 2002-2018 vs each team's average win total for the preceding four years.

  • Now we're seeing a much clearer trend in the data in Figure 6. The data grouping is much tighter and follows the trendline you'd expect to see: top-left downwards toward the bottom-right (i.e. better recruiting with more wins).
  • If a team averages more than 10 wins over a four year period, their next recruiting class is almost guaranteed to be in the Top 20. A major exception to this is Virginia Tech, which finished inside the Top 20 only once out of six years with a 4-year average of 10+ wins. This illustrates that that program has some sort of major disadvantage with regards to recruiting compared to other high-achieving teams.
  • Something jumped out at me that I didn't expect when looking at this chart, and it has to do with team colors. If your team has a single primary color and that color is any kind of shade of blue, it will almost never average more than 8 wins over a 4-year period. Duke average 8.3 wins from 2012-2015, and UNC averaged 8 wins for the last two years of the period 2012-2016. That's it. Also, of the 'blue' teams only UNC has ever had a Top 20 recruiting class.

Figure 7. 247 Composite Team Recruiting Rankings (Top 25 only) for each year over the period 2002-2018 vs each team's average win total for the preceding four years.

  • Again, notice the much tighter distribution of data points in Figure 7 compared to Figures 3 and 5.
  • Three teams have gotten their 4-year win average to 12 or above: Alabama, Florida State, and Clemson. Of these Clemson has underachieved relative to the other two in recruiting rankings. Clemson finished 16th in 2017 and then 7th in the most recent recruiting class after averaging more that 12.3 wins in the those years. The lowest the other two have ranked when averaging more than 12 wins is 6th (Alabama, 2018), with most rankings ending up as either 1st (Alabama) or 3rd (FSU).
  • In Figure 3 you can see that many teams have been able to recruit at a Top 25-level after having only 4 or fewer wins the previous season. Figure 7 shows, however, that it is almost impossible to pull that off if you only average less than 4 wins over a 4-year period preceding the recruiting class.
  • North Carolina managed a 10th ranked finish in recruiting in 2007 despite averaging only 4 wins. Weird. Ole Miss and Tennessee also managed Top 10 recruiting rankings while averaging less than 6 wins. Again, we'll talk more about these classes a little later.

Extrapolations from Wins and Recruiting Data

So far we've seen recruiting rankings compared against three different views of prior wins: wins in the season immediately preceding the recruiting class, wins in the season two years prior to the class, and average wins for the four season prior to the recruiting class. In Figure 8 you can see the trendline for each of the three separate datasets.

Figure 8. The trendlines for each of the full datasets in Figures 2, 4, and 6.

  • The first thing to notice is that the trendlines for the 'Preceding Year' and '2 Years Prior' datasets are nearly identical. These trendlines project for every 1 win either the previous season or two season back, a team's recruiting ranking would be expected to improve by 3.9 spots. You can write this equation to roughly predict a team's recruiting ranking based on a given win total:

Recruiting Ranking = 60 - 3.9 * # of Wins

  • That equation won't be very good, however, as the R-squared number for those trendlines is fairly low. R-squared is a measurement that shows how closely data fits a particular model. In this case the R-squared values for the 'Preceding Year' and '2 Years Prior' trendlines are 22.8% and 24.2%, respectively. In other words, only about 23.5% of recruiting rankings will fit a linear model based on wins from recent single seasons. As we visually saw in the data earlier though, the linear model fits the '2 Years Prior' dataset slightly better than it does the 'Preceding Year' dataset.
  • The correlation coefficient for the 'Preceding Year' and '2 Years Prior' datasets are -0.48 and -0.49, respectively. This indicates a moderate linear relationship.
  • The R-squared value for the linear model of the '4-Year Average' dataset is much better at 35.0%. This certainly jives with a visual inspection of Figure 6 compared to either Figures 2 or 4. The slope of the line here is much steeper, with every increase in 4-year average of 1 win corresponding to a 6.1 improvement in recruiting ranking. The equation for this line can be written as:

Recruiting Ranking = 76 - 6.1 * 4-Year Win Average

  • The steeper slope of the trendline for the '4-Year Average' dataset isn't all that surprising when you consider that improving your 4-year win average by 1 represents a total of 4 additional wins over that period. This corresponds to roughly winning 8% more of your total games.
  • The correlation coefficient for the '4-Year Average' dataset is also much stronger at -0.59, which represents a moderately strong linear relationship.

Improvement in Wins and Recruiting

OK, so now we've shown that there is a (weak) correlation between wins and recruiting rankings, but what about the effect of improvement in wins? Does the excitement generated by a drastic increase in the win column spill over to the recruiting trail?As we see in the Figure 9, the answer to that question is: Nope!

Figure 9. Year-over-year change in Recruiting Rankings from 2003-2018 compared against year-over-year change in win totals in the preceding year.

  • If rising (or falling) win totals had an effect on recruiting ranking, the logical expectation for what you'd see in Figure 9 would be a diagonal line moving from the bottom-left up towards the top-right of the chart. The chart above doesn't show anything like that, however.
  • Given the symmetrical clustering around the origin in the chart, what is shown is that change in recruiting ranking appears to move independently of change in preceding year win totals.
  • This is backed-up by the very low correlation coefficient of this dataset of -0.10, indicating almost no linear relationship whatsoever.
  • So, combining this with what we saw above, we can say that while wins themselves are somewhat correlated with future recruiting performance, change in win totals in themselves do not correspond to change in recruiting rankings.

Top 10 Recruiting with 6 Wins or Less the Previous Year

I've called attention a few times above to the relatively rare phenomenon of teams achieving elite recruiting results despite limited success on the field. Given the small number of teams involved I was able to add labels to the chart in Figure 10. For further clarity I've listed the teams that have managed to pull in Top 10 recruiting classes after winning 6 or less games in the preceding year:

  • 2004 Texas A&M - #10 after 4 wins
  • 2007 North Carolina - #10 after 3 wins
  • 2008 Miami - #1 after 5 wins
  • 2009 Tennessee - #8 after 5 wins
  • 2011 Clemson - #10 after 6 wins
  • 2011 Georgia - #7 after 6 wins
  • 2012 Miami - #10 after 6 wins
  • 2014 Florida - #9 after 4 wins
  • 2014 Tennessee - #7 after 5 wins

Figure 10. 247 Composite Team Recruiting Rankings (Top 10 only) during the period 2002-2018 for teams that finished with less than 6 wins in the preceding year.

Of the 9 teams that have achieved this feat, a surprisingly (or maybe unsurprisingly???) large number of them were caught up in varying degrees of scandal afterwards.

One team that just missed being included in the list above by squeezing out 7 wins immediately prior to finishing #8 in the recruiting rankings is Ole Miss. The sordid history there has been well documented, but if you have the time you should definitely check out Steven Godfrey's long-form piece on the subject.

Given the assumed prevalence of cheating in college football and relative rarity of any team actually ever getting caught, the fact that 4 out of the 9 teams listed above were later involved in scandals of varying degrees is telling. In a lot of cases, though, appearance on this list could be the result of a traditional recruiting powerhouse simply having a bad year (Geogia in 2011, Will Muschamp's 2014 Florida team), or former blue-blood program getting a jolt of energy from a new head coach in his first full recruiting class during a rebuild (Butch Jones, Tennessee 2014).

A Closer Look at the Rivals

The charts above can be hard to decipher given the 475 data points (on Figures 2, 4, and 6). I'm obviously interested to see look at what happens to all teams in the SEC and ACC, but what I'm most interested to see is how the Gamecocks compare to the two main rivals, Clemson and Georgia.

Figure 11. 247 Composite Team Recruiting Rankings for South Carolina, Georgia, and Clemson each year over the period 2002-2018 vs each team's win total for the preceding year.

  • As I have noted previously, South Carolina and Georgia have been very consistent this century in terms of recruiting ranking. Georgia just tends to recruiting about 14 spots better than SC on average. Georgia's average ranking is 6.9, with the standard deviation being an incredibly low 2.8.  The Gamecocks' average is 20.7, with a relatively low standard deviation of 7.8.
  • On the other hand, Clemson recruiting has been all over the place the last few 17 cycles. The average ranking is less than 1 better than SC's at 19.8, but the standard deviation is large at 11.4. Clemson has also demonstrated a lot more variance in terms of number of wins, driven by their rapid improvement in that area since the beginning of this decade.

Figure 12. 247 Composite Team Recruiting Rankings for South Carolina, Georgia, and Clemson each year over the period 2002-2018 vs each team's average win total for the preceding four years.

  • Look at how tight that Georgia data cluster is in Figure 12. None of the 28 teams viewed for this analysis came anywhere close to the same level of consistency in both recruiting and wins as Georgia has this century.

Average Results by Team and Conference

As a finale I thought it would be interesting to look at how each team in the SEC and ACC have performed on average for the past 17 years. In Figure 13 I've plotted that data, along with average values and trendlines for each conference.

Figure 13. Average wins over the period 2001 - 2017 and average recruiting ranking over the period 2002 - 2018 for each team and conference.

  • ACC teams have averaged 7.15 wins from 2001 to 2017, and had a average recruiting ranking of 39.1 from 2002 to 2018. SEC teams averaged 7.57 wins and a recruiting ranking of 24.0.
  • Consider that the ACC and SEC both play 113 conference games each year (14 teams, 8 regular season conference games, 1 conference championship game). This means that each conference starts with a baseline average of 4.04 wins per team just from conference games. The SEC has therefore done a good bit better than the ACC in non-conference games and bowl games, with each team netting an average of 0.42 more wins per year. In total, SEC teams have accrued 1,801 wins over the past 17 years, while the ACC has exactly 100 less at 1,701.
  • The head-to-head record between the two conferences over this period is 86-65 in favor of the SEC. This is a win percentage of 57.0%.
  • The bottom-right of the chart is dominated by 4 SEC teams (Alabama, LSU, Georgia, and Florida), with an intrusion by only one ACC team (Florida State).
  • The most representative ACC team based on averages is Pittsburg. For the SEC, the most representative team is .......... South Carolina.
  • Looking at the data sets you can almost see two distinct groups, with the ACC teams forming a downward sloping line above the one formed by the SEC teams. The trendlines added for each conference highlight this, and you can see that these are almost parallel with one another. Overall data for the ACC teams tend to be above and to the left of that for the SEC teams.
  • From just looking at the conference averages you can see that SEC teams recruit at a level about 15 spots above ACC teams. Looking at the trendlines you can also say that an average ACC team will win about 1.8 games more per year than an SEC team that recruits at the same level.

Conclusions

I started writing this post way back at the end of June, but as I progressed I kept getting ideas for things to add and other ideas about how to look at the original data. As a result it has grown much longer and more random than I initially intended. There are a lot of words written out above, but here are the main points I think you can draw from the various views of the data:

  • Wins from the preceding year and two years back are nearly indistinguishable in terms of ability to predict recruiting ranking. Both show moderate linear correlation with recruiting, although a linear regression of the data only fits a little less than 25% of the results.
  • The 4-year win average is much better at predicting recruiting rankings, as the data shows a moderately high correlation. A linear regression fits 35% of results.
  • A team that averages less than 4 wins over a 4-year period will not be able to recruit at a Top 25 level. Conversely, a team that averages over 10 wins will almost certainly recruit at a Top 20 level.
  • If a team has a single primary color and that color is blue, the team is most likely terrible.
  • If a team improves (or reduces) it's win total, the magnitude of the improvement / reduction itself doesn't appear to have any impact on any change in the team's recruiting ranking.
  • A team that wins less than 6 games and then hauls in a a Top 10 recruiting class will find itself mixed in with some disreputable company.
  • The SEC has dominated the ACC since 2001 not only in head-to-head record, but also recruiting and total wins.

See the table below for the details of the dataset used to create the above charts.