:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/48737425/usa-today-9024544.0.jpg)
Right before the 2015 season, I wrote a post at Football Study Hall exploring a different way to measure returning experience that goes beyond "How many starters do you return?"
We use returning starters because nothing else is particularly available. It is fine as a snapshot, but we know one team's six returning offensive starters aren't another's. What about go-to guys? Returning backups? And quarterbacks are worth more than other starters, right?
You can see some people drawing reference to percent of returning yards/tackles/etc., but it's hard to find concrete data.
Of course, I do compile all of this data, don't I? ... Why don't I weaponize this data for use in projections?
My goal was to begin compiling "percentage returning" data for every level of an offense and defense -- passing yards/attempts/completions, rushing attempts/yards, receiving targets/receptions/yards, offensive line starts (because that's all we have), and tackles/TFLs/pass breakups at each level of the defense.
Once I had this, I could tinker to see how much each category affects a team's offensive or defensive improvement. In theory, this could allow me to create a "percentage of offense/defense returning" figure that could dwarf the effectiveness of just data on returning starters.
With just one year at my disposal (2014) when I wrote the original post, I wasn't able to draw any lasting conclusions, but the potential was obvious. I put together a rough formula that posited that teams like North Carolina (87 percent of production returning), Ohio (85 percent), and Temple (83 percent) were among the most experienced in the country, then watched as those teams improved from a combined 18-9 to a combined 29-12. (Others near the top didn't improve nearly that much.)
Meanwhile, the five that returned less than 40 percent of their production -- Kansas, UTEP, Wyoming, UCF, and UL-Lafayette -- fell from a combined 32-31 to 11-49.
At the end of 2015, I was able to add a second year of data. I have 2013 queued up and ready to go. But the conclusions are becoming clear.
On Monday, I will post initial 2016 S&P+ projections and begin my 2016 offseason preview series. Here's the returning production data I will be using along with these recruiting rankings and other factors. The projected points per game in the table below simply refers to points on the scoreboard.
| O returning | Proj. O PPG change |
D returning |
Proj. D PPG change |
Overall returning |
Overall proj. PPG change |
Rank | |
| LSU | 96% | +5.0 | 88% | -4.4 | 92% | +9.4 | 1 |
| UCF | 99% | +5.4 | 84% | -3.6 | 92% | +9.0 | 2 |
| Syracuse | 100% | +5.5 | 80% | -3.0 | 90% | +8.5 | 3 |
| Kent State | 94% | +4.7 | 83% | -3.5 | 89% | +8.2 | 4 |
| Charlotte | 92% | +4.5 | 82% | -3.3 | 87% | +7.8 | 5 |
| Rutgers | 81% | +2.9 | 90% | -4.6 | 85% | +7.5 | 6 |
| Louisville | 98% | +5.2 | 77% | -2.3 | 87% | +7.5 | 7 |
| Kansas | 81% | +2.9 | 85% | -3.9 | 83% | +6.8 | 8 |
| Wake Forest | 89% | +4.0 | 79% | -2.7 | 84% | +6.7 | 9 |
| Ball State | 78% | +2.5 | 87% | -4.1 | 82% | +6.6 | 10 |
| Nevada | 95% | +4.9 | 73% | -1.7 | 84% | +6.6 | 11 |
| Tennessee | 84% | +3.3 | 78% | -2.6 | 81% | +5.9 | 12 |
| UConn | 94% | +4.7 | 70% | -1.1 | 82% | +5.8 | 13 |
| Texas | 79% | +2.7 | 81% | -3.1 | 80% | +5.8 | 14 |
| BC | 92% | +4.4 | 71% | -1.3 | 81% | +5.7 | 15 |
| Colorado | 81% | +2.9 | 78% | -2.6 | 80% | +5.5 | 16 |
| Army | 71% | +1.6 | 85% | -3.8 | 78% | +5.4 | 17 |
| SMU | 97% | +5.1 | 65% | -0.3 | 81% | +5.4 | 18 |
| New Mexico | 76% | +2.3 | 80% | -3.0 | 78% | +5.3 | 19 |
| South Florida | 89% | +4.0 | 70% | -1.2 | 79% | +5.2 | 20 |
| Idaho | 82% | +3.0 | 76% | -2.1 | 79% | +5.1 | 21 |
| Miami (Ohio) | 84% | +3.3 | 74% | -1.8 | 79% | +5.1 | 22 |
| WSU | 87% | +3.7 | 71% | -1.4 | 79% | +5.1 | 23 |
| NMSU | 88% | +3.8 | 71% | -1.3 | 79% | +5.1 | 24 |
| Nebraska | 94% | +4.7 | 63% | +0.1 | 78% | +4.6 | 25 |
| EMU | 78% | +2.6 | 75% | -2.0 | 77% | +4.6 | 26 |
| Washington | 72% | +1.7 | 79% | -2.8 | 76% | +4.5 | 27 |
| UCLA | 61% | +0.2 | 88% | -4.3 | 75% | +4.5 | 28 |
| Georgia | 75% | +2.1 | 76% | -2.3 | 76% | +4.4 | 29 |
| Duke | 80% | +2.7 | 73% | -1.6 | 76% | +4.3 | 30 |
| App State | 72% | +1.7 | 78% | -2.6 | 75% | +4.3 | 31 |
| Virginia Tech | 66% | +0.9 | 82% | -3.3 | 74% | +4.2 | 32 |
| Oregon State | 72% | +1.6 | 78% | -2.5 | 75% | +4.1 | 33 |
| Fresno State | 89% | +4.1 | 62% | +0.2 | 76% | +3.9 | 34 |
| Missouri | 82% | +3.0 | 69% | -0.9 | 75% | +3.9 | 35 |
| Texas Tech | 76% | +2.2 | 72% | -1.6 | 74% | +3.8 | 36 |
| OK State | 79% | +2.6 | 69% | -0.9 | 74% | +3.5 | 37 |
| NIU | 85% | +3.5 | 63% | +0.1 | 74% | +3.4 | 38 |
| Wyoming | 60% | +0.0 | 83% | -3.4 | 71% | +3.4 | 39 |
| Vanderbilt | 59% | -0.1 | 83% | -3.5 | 71% | +3.3 | 40 |
| CMU | 80% | +2.7 | 67% | -0.6 | 73% | +3.3 | 41 |
| Pittsburgh | 72% | +1.6 | 72% | -1.6 | 72% | +3.2 | 42 |
| Iowa | 72% | +1.7 | 72% | -1.4 | 72% | +3.1 | 43 |
| ODU | 65% | +0.8 | 76% | -2.2 | 71% | +3.0 | 44 |
| Virginia | 80% | +2.8 | 64% | -0.1 | 72% | +2.9 | 45 |
| ECU | 85% | +3.5 | 60% | +0.6 | 73% | +2.9 | 46 |
| UTEP | 89% | +4.0 | 56% | +1.2 | 73% | +2.8 | 47 |
| BYU | 67% | +1.0 | 73% | -1.7 | 70% | +2.7 | 48 |
| Cincinnati | 55% | -0.6 | 82% | -3.2 | 69% | +2.6 | 49 |
| WMU | 79% | +2.6 | 62% | +0.2 | 70% | +2.4 | 50 |
| Tulsa | 73% | +1.7 | 67% | -0.7 | 70% | +2.4 | 51 |
| Kansas State | 77% | +2.4 | 62% | +0.2 | 70% | +2.2 | 52 |
| Iowa State | 65% | +0.7 | 71% | -1.4 | 68% | +2.1 | 53 |
| Miami | 78% | +2.4 | 62% | +0.3 | 70% | +2.1 | 54 |
| Oklahoma | 72% | +1.7 | 66% | -0.4 | 69% | +2.1 | 55 |
| Baylor | 71% | +1.5 | 66% | -0.5 | 69% | +2.0 | 56 |
| Indiana | 47% | -1.8 | 84% | -3.7 | 66% | +1.9 | 57 |
| Boise State | 87% | +3.7 | 53% | +1.9 | 70% | +1.8 | 58 |
| Rice | 51% | -1.3 | 80% | -3.0 | 66% | +1.7 | 59 |
| Hawaii | 66% | +0.8 | 69% | -0.9 | 67% | +1.7 | 60 |
| Clemson | 90% | +4.1 | 50% | +2.4 | 70% | +1.7 | 61 |
| Southern Miss | 65% | +0.7 | 69% | -1.0 | 67% | +1.7 | 62 |
| SDSU | 52% | -1.1 | 79% | -2.8 | 66% | +1.7 | 63 |
| USC | 54% | -0.8 | 77% | -2.5 | 66% | +1.6 | 64 |
| Maryland | 93% | +4.6 | 46% | +3.1 | 70% | +1.5 | 65 |
| MTSU | 80% | +2.8 | 55% | +1.4 | 68% | +1.4 | 66 |
| Texas A&M | 59% | -0.2 | 72% | -1.6 | 65% | +1.4 | 67 |
| UL-Lafayette | 54% | -0.8 | 75% | -2.1 | 65% | +1.3 | 68 |
| Oregon | 49% | -1.5 | 79% | -2.8 | 64% | +1.3 | 69 |
| Georgia Tech | 97% | +5.1 | 42% | +3.8 | 70% | +1.3 | 70 |
| GA Southern | 89% | +3.9 | 48% | +2.6 | 69% | +1.3 | 71 |
| Purdue | 68% | +1.1 | 65% | -0.2 | 66% | +1.3 | 72 |
| Arkansas | 37% | -3.2 | 89% | -4.5 | 63% | +1.3 | 73 |
| North Carolina | 50% | -1.3 | 78% | -2.5 | 64% | +1.2 | 74 |
| Tulane | 54% | -0.9 | 75% | -2.0 | 64% | +1.2 | 75 |
| Michigan | 53% | -0.9 | 73% | -1.7 | 63% | +0.8 | 76 |
| Minnesota | 84% | +3.3 | 48% | +2.6 | 66% | +0.7 | 77 |
| Arizona | 61% | +0.1 | 65% | -0.3 | 63% | +0.4 | 78 |
| UL-Monroe | 78% | +2.5 | 51% | +2.2 | 65% | +0.3 | 79 |
| Air Force | 34% | -3.6 | 85% | -3.9 | 60% | +0.3 | 80 |
| South Carolina | 61% | +0.2 | 63% | 0.0 | 62% | +0.2 | 81 |
| Northwestern | 68% | +1.1 | 58% | +0.9 | 63% | +0.2 | 82 |
| Ole Miss | 74% | +1.9 | 54% | +1.7 | 64% | +0.2 | 83 |
| Florida Int'l | 80% | +2.7 | 48% | +2.6 | 64% | +0.1 | 84 |
| Auburn | 64% | +0.6 | 61% | +0.5 | 62% | +0.1 | 85 |
| Temple | 73% | +1.8 | 53% | +1.8 | 63% | 0.0 | 86 |
| NC State | 50% | -1.4 | 70% | -1.2 | 60% | -0.1 | 87 |
| Akron | 72% | +1.6 | 53% | +1.8 | 63% | -0.2 | 88 |
| Marshall | 72% | +1.6 | 53% | +1.8 | 62% | -0.2 | 89 |
| Kentucky | 62% | +0.3 | 61% | +0.5 | 62% | -0.2 | 90 |
| Florida State | 77% | +2.3 | 49% | +2.5 | 63% | -0.2 | 91 |
| SJSU | 69% | +1.2 | 55% | +1.5 | 62% | -0.3 | 92 |
| North Texas | 42% | -2.5 | 76% | -2.2 | 59% | -0.4 | 93 |
| Florida | 63% | +0.4 | 59% | +0.8 | 61% | -0.4 | 94 |
| UTSA | 61% | +0.2 | 58% | +0.9 | 60% | -0.7 | 95 |
| West Virginia | 86% | +3.6 | 38% | +4.5 | 62% | -0.9 | 96 |
| Penn State | 53% | -1.0 | 63% | +0.0 | 58% | -1.0 | 97 |
| FAU | 54% | -0.9 | 62% | +0.3 | 58% | -1.2 | 98 |
| Utah | 35% | -3.5 | 76% | -2.2 | 55% | -1.3 | 99 |
| Georgia State | 45% | -2.1 | 68% | -0.7 | 56% | -1.3 | 100 |
| Miss. State | 44% | -2.2 | 69% | -0.9 | 56% | -1.3 | 101 |
| Texas State | 52% | -1.2 | 62% | +0.3 | 57% | -1.5 | 102 |
| Houston | 72% | +1.7 | 44% | +3.3 | 58% | -1.6 | 103 |
| Alabama | 45% | -2.1 | 66% | -0.5 | 55% | -1.6 | 104 |
| Memphis | 40% | -2.7 | 68% | -0.8 | 54% | -1.9 | 105 |
| UNLV | 56% | -0.5 | 55% | +1.5 | 56% | -2.0 | 106 |
| Ark. State | 37% | -3.2 | 69% | -1.0 | 53% | -2.1 | 107 |
| Troy | 57% | -0.4 | 52% | +1.9 | 55% | -2.3 | 108 |
| Illinois | 74% | +2.0 | 38% | +4.6 | 56% | -2.6 | 109 |
| Notre Dame | 60% | +0.1 | 47% | +3.0 | 54% | -2.9 | 110 |
| USA | 34% | -3.6 | 64% | -0.1 | 49% | -3.5 | 111 |
| TCU | 29% | -4.3 | 68% | -0.8 | 48% | -3.5 | 112 |
| Ohio | 64% | +0.5 | 40% | +4.1 | 52% | -3.6 | 113 |
| Buffalo | 18% | -5.8 | 73% | -1.7 | 46% | -4.1 | 114 |
| Stanford | 33% | -3.7 | 61% | +0.5 | 47% | -4.2 | 115 |
| Utah State | 57% | -0.4 | 41% | +3.9 | 49% | -4.3 | 116 |
| WKU | 34% | -3.6 | 59% | +0.7 | 47% | -4.3 | 117 |
| CSU | 58% | -0.3 | 40% | +4.2 | 49% | -4.5 | 118 |
| BGSU | 29% | -4.2 | 62% | +0.3 | 46% | -4.5 | 119 |
| Mich. State | 26% | -4.6 | 64% | 0.0 | 45% | -4.6 | 120 |
| Navy | 34% | -3.6 | 56% | +1.4 | 45% | -5.0 | 121 |
| Wisconsin | 34% | -3.6 | 54% | +1.7 | 44% | -5.3 | 122 |
| Toledo | 44% | -2.2 | 43% | +3.6 | 43% | -5.8 | 123 |
| California | 15% | -6.2 | 62% | +0.2 | 39% | -6.4 | 124 |
| LA Tech | 42% | -2.5 | 37% | +4.7 | 39% | -7.2 | 125 |
| Arizona State | 27% | -4.5 | 43% | +3.7 | 35% | -8.2 | 126 |
| Ohio State | 22% | -5.2 | 36% | +4.7 | 29% | -9.9 | 127 |
| UMass | 15% | -6.3 | 34% | +5.2 | 24% | -11.5 | 128 |
No pressure, Les.
LSU was ninth in S&P+ last season, ranks eighth in two-year recruiting, and is first in returning experience. I don't know, do you figure my projections might have the Tigers pretty high next week?
LSU fell into a three-week funk last November, which colored our perceptions but only did so much to temper their numbers. The Tigers otherwise played at the 81st percentile or higher in every game and are scheduled to return last year's quarterback, every rusher (including that Leonard Fournette guy), 10 of their top 11 receiving targets, every defensive lineman, and nine of 10 defensive backs. Last year's freshmen and sophomores are this year's sophomores and juniors.
There are still questions -- Can Brandon Harris develop into a more efficient passer? Do experience and a new defensive coordinator shore up a leaky run defense? -- but on paper, the pieces are in place for the Tigers in a way they haven't been in years.
Which experience matters?
I didn't say anything about LSU's offensive line or linebackers. The Tigers do return three OL starters, and linebacker Kendell Beckwith elected to return for his senior season. But from a statistical perspective, it appears experience in these two units means very little.
With a couple years of data, here are the correlations between a percentage returning category and change in Off. S&P+ (the higher, the more correlated returning experience is with production):
- Receiving yards returning: 0.285
- Passing yards returning: 0.264
- Rushing yards returning: 0.079
- Career offensive line starts returning: 0.015
And here are the correlations between some returnee categories and Def. S&P+:
- Passes broken up returning (overall): -0.440
- Passes broken up returning (DBs): -0.404
- Tackles returning (overall): -0.388
- Tackles returning (DBs): -0.378
- Sacks returning (DLs): -0.194
- Passes broken up returning (DLs): -0.161
- Tackles returning (LBs): -0.161
Offensive line experience, as calculated by career starts returning, has almost no impact on a team's Off. S&P+. That is rather mind-blowing. Now, that could be because career starts aren't the best measure to use. I will look into using solely the previous year's total -- percentage of 2015 OL starts returning, in other words -- and putting a cap on career starts. As it stands, losing one three-year starter and returning four one-year starters means you've barely returned 50 percent of your career starts. That might not be the best way to measure things.
For now, however, the correlations are null. And for that reason, this initial correlation ignores offensive line starts*. That feels like a strange thing to do, but if a measure has no impact, there is no reason to use it.
When these projections are finalized after spring football, I'll hopefully have data for stronger conclusions.
* I don't yet have all the offensive line start data compiled anyway, so this is convenient.
As with recruiting, experience matters more on defense.
The squared correlation between a team's defensive percentage returning and Def. S&P+ change is 22.1 percent, meaning this change in experience can account for 22 percent of a team's change in defensive quality.
On offense, it's only 13.8 percent.
As you see in the table above, experience can be a little more powerful on defense, at least at the extreme end. Returning 34 percent of your offense, like Air Force does, means regression of about 3.6 points per game, while returning 34 percent of your defense, like UMass does, means 5.2 points. On the other end, Clemson returning 90 percent of its offense benefits it by 4.1 points; Rutgers returning 90 percent of its defense means 4.6.
Experience in the secondary is worth more than experience in the front seven. And it appears that the skill of being able to either pick off or bat down passes is far more difficult to replace than other skills.
Do you have Ohio State and Stanford in your top 10?
Because if so ... well ... consider this a red flag. The Cardinal are under the 50 percent bar, and Buckeyes are one of five teams projected to return less than 40 percent of their production. That meant doom across the board last year, and while both of these programs could fall a couple of notches and still be quite good, you might not want to expect too much. That might serve as an alarm bell to many far-too-early prognosticators.
Two other teams on the wrong side: Cal and Arizona State.
Cal's easy to figure out -- the Golden Bears are a pass-first team that must replace their quarterback (soon to become a high draft pick) and top six receiving targets. And as we see above, quarterback and receiver experience are worth far more than running back or offensive line experience. (Losing two starters in the defensive backfield doesn't help.)
Arizona State is coming off of a terribly disappointing 6-7 campaign, and head coach Todd Graham has been tasked with replacing longtime offensive coordinator Mike Norvell (who took the Memphis head coaching job), tight ends coach Chip Long (who followed Norvell), and defensive line coach Jackie Shipp (who left for the same position at Missouri). The Sun Devils must now replace their quarterback, their top two receivers, and four of their top five defensive backs. Yikes.
Approaching returning talent like this is exciting. It has already led to projections more strongly correlated to improvement and regression, and there are still further matters to tackle -- how to handle offensive line experience, for instance. I feel comfortable enough with this data to use them in the upcoming projections and previews, but hopefully I'll get a little bit further with this by the next round of projections.
What stands out to you in the table above?
(Note: This post was updated to account for some roster errors. Michigan, Notre Dame, and Stanford do not list redshirts, so some of the players listed as seniors are juniors, juniors are sophomores, etc. It's annoying, and it manages to trip me up every damn year. Granted, that's partially on me. But I will still complain about it.)