clock menu more-arrow no yes

Filed under:

Let’s see what Italian economists have to say about college football recruiting rankings

New, 10 comments

Pareto principle, meet S&P+ projections.

Wikimedia Commons

1.

We’ll begin this class by quoting Wikipedia:

The Pareto principle (also known as the 80/20 rule, the law of the vital few, or the principle of factor sparsity) states that, for many events, roughly 80% of the effects come from 20% of the causes. Management consultant Joseph M. Juran suggested the principle and named it after Italian economist Vilfredo Pareto, who noted the 80/20 connection while at the University of Lausanne in 1896, as published in his first work, Cours d’économie politique. Essentially, Pareto showed that approximately 80% of the land in Italy was owned by 20% of the population.

2.

For a while now, I’ve had the word “Pareto” on my long-term to-do list. I’ve wanted to see if this principle holds true in football recruiting. I can tie anything in the world back to football.

3.

I pulled in the roster/recruiting data from my 2018 previews and basically ranked the top 17 players on the roster, per the 247Sports Composite (because 17 is 20% of 85 scholarships). I also made the following adjustments to those Composite ratings:

  • If you were unrated in the Composite, I assigned a 0.6800 rating. The lowest you’ll find for FBS recruits in general is a 0.7000, so that seemed to make sense for walk-ons I deemed worth listing in the previews. But this doesn’t really matter because not even service academies will have unrated recruits in their top 17. I just felt like sharing.
  • I added a multiplier based on which unit you play on so that each unit carries the same average. (For 2018 rosters, the average Composite rating for WR/TEs was 0.8245, DLs 0.8231, QBs 0.8231, DBs 0.8192, LBs 0.8188, RBs 0.8188, and OLs 0.8184. So I adjusted it so that they’re all 0.8208 — the midpoint.)
  • I also made an adjustment based on whether you’re on offense or defense. Why? Because recruiting rankings for defenders always have stronger correlations to performance than rankings for offensive players. So based on correlations, your rating is multiplied by 1.04 if you’re a defensive player and 0.96 if you’re an offensive player. That means that quite a few defensive players ended up with an Adjusted Composite rating over 1.0000, which is also fun.

4.

The correlation between the 2-year recruiting rankings I used for 2018 S&P+ projections and the current year-end 2018 S&P+ rankings was a solid 0.609. That’s a strong correlation, and it’s why I use recruiting rankings in my projections.

5.

The correlation between these new Pareto recruiting rankings and 2018 S&P+: 0.640. Better.

6.

Take out the massively underachieving USC and FSU, and it rises to 0.661. Jerks.

7.

I also looked at the the average 247 rating for the top 13 non-freshmen on your roster, since even star freshmen rarely play massive roles, and since 13 is about 20 percent of your non-freshman roster on average. The correlation between this top-13 average and 2018 S&P+: 0.633. Still better than the old way but not quite as good (or easy!) as just taking the overall top 17.

8.

Taking out the mid-majors, who are always a different story for obvious reasons, the top 12 P5 teams in S&P+ this year were Alabama, Clemson, Georgia, Oklahoma, Michigan, Notre Dame, Ohio State, Washington, Penn State, Mississippi State, Florida, and LSU. All were in the top 22 of these Pareto rankings, and 10 of them were in the top 15.

9.

The first major P5 outlier: MIZZOU, BABY. 13th-best P5 team per S&P+, 37th in the Pareto rankings.

10.

Why might it be more accurate to do it this way? For one thing, it’s serving a different purpose than the two-year rating. I use two-year averages in the S&P+ projections because it asks a follow-up question to “What production are they returning?” It basically gets at “How good are your new starters?” Using five-year recruiting averages correlates slightly better to S&P+ (0.619), but that, too, is looking more at roster talent than the talent of new contributors.

11.

Still, this Pareto average appears to be even better than the five-year average. Why? For one thing, it accounts for transfers, both in and out. If you sign blue-chippers and then lose them to transfers (looking at you, Auburn), you don’t get propped up as much. If you bring in a lot of transfers (hello, Houston and WVU), then doing this will more accurately read the talent on your roster.

12.

You want to see the data? HERE’S THE DATA.

Pareto recruiting rankings

Team Pareto top 17 avg Rk 5-yr rec. pctile Rk 2018 S&P+ (to date) Rk
Team Pareto top 17 avg Rk 5-yr rec. pctile Rk 2018 S&P+ (to date) Rk
Alabama 1.0142 3 0.9889 1 29.67 1
Clemson 1.0006 6 0.9498 8 27.89 2
Georgia 1.0019 5 0.9754 3 25.89 3
Oklahoma 0.9844 10 0.9249 13 22.15 4
Michigan 0.9846 9 0.8749 19 21.28 5
Notre Dame 0.9662 15 0.9384 9 20.58 6
Ohio State 1.0194 1 0.9842 2 19.82 7
Central Florida 0.8969 52 0.5135 65 19.60 8
Washington 0.9529 20 0.8230 23 17.03 9
Fresno State 0.8613 87 0.3013 85 16.22 10
Penn State 0.9721 13 0.9066 15 16.09 11
Mississippi State 0.9473 22 0.7943 26 15.61 12
Appalachian State 0.8359 118 0.2006 108 15.49 13
Florida 0.9746 12 0.9251 11 15.12 14
LSU 1.0073 4 0.9686 6 14.78 15
Missouri 0.9175 37 0.6746 40 14.38 16
Utah 0.9137 41 0.6669 42 14.36 17
Auburn 0.9654 16 0.9540 7 12.80 18
Texas A&M 0.9643 17 0.9262 10 12.69 19
West Virginia 0.9229 32 0.6870 39 12.63 20
Utah State 0.8479 103 0.1891 113 12.42 21
Miami-FL 0.9585 18 0.8985 17 12.10 22
Memphis 0.8841 67 0.4096 73 12.02 23
Wisconsin 0.9120 43 0.7279 34 11.41 24
Washington State 0.8872 63 0.5793 53 10.88 25
Iowa 0.9107 45 0.6131 47 10.79 26
Stanford 0.9554 19 0.8750 18 10.61 27
Oklahoma State 0.9137 40 0.7072 37 10.36 28
Boise State 0.8824 69 0.5166 63 9.97 29
NC State 0.9201 35 0.6999 38 9.90 30
North Texas 0.8465 104 0.2049 105 9.76 31
Cincinnati 0.8916 57 0.5010 66 9.01 32
South Carolina 0.9431 24 0.8612 21 8.76 33
Temple 0.8753 74 0.3848 76 8.07 34
Michigan State 0.9346 27 0.7966 25 7.71 35
Texas 0.9911 8 0.9250 12 7.58 36
Purdue 0.8752 75 0.4765 69 6.70 37
San Diego State 0.8690 80 0.3991 75 6.64 38
USC 1.0144 2 0.9743 4 6.55 39
Kentucky 0.9204 34 0.7528 33 6.55 40
Houston 0.9083 46 0.4597 70 6.24 41
Virginia 0.8977 50 0.5906 51 6.21 42
Syracuse 0.8891 60 0.5665 55 6.09 43
Texas Tech 0.8862 66 0.6313 45 6.06 44
Ohio 0.8337 121 0.1951 110 5.59 45
Troy 0.8625 84 0.2104 103 5.36 46
Oregon 0.9510 21 0.8674 20 5.23 47
Toledo 0.8675 81 0.3368 80 4.58 48
Marshall 0.8751 76 0.4209 72 4.38 49
Arkansas State 0.8560 93 0.2623 91 4.22 50
Buffalo 0.8441 106 0.1504 120 4.20 51
Iowa State 0.8902 59 0.5614 56 3.98 52
UAB 0.8587 90 0.2012 107 3.73 53
Arizona State 0.9349 26 0.7907 28 3.45 54
Nebraska 0.9343 29 0.7986 24 3.18 55
BYU 0.9068 47 0.4928 68 3.07 56
TCU 0.9159 38 0.7535 32 2.99 57
Minnesota 0.8952 53 0.5931 50 2.93 58
Georgia Southern 0.8568 92 0.2754 88 2.88 59
Florida Atlantic 0.8808 72 0.3585 78 2.86 60
Middle Tennessee 0.8508 98 0.2594 92 2.78 61
Miami-OH 0.8503 99 0.2189 100 2.40 62
Eastern Michigan 0.8383 112 0.1457 123 2.01 63
California 0.8972 51 0.6384 44 1.57 64
Northern Illinois 0.8494 101 0.2223 97 1.56 65
Duke 0.9109 44 0.6301 46 1.39 66
Pittsburgh 0.9190 36 0.6671 41 1.39 67
Vanderbilt 0.9134 42 0.6104 48 1.38 68
Virginia Tech 0.9345 28 0.7757 30 1.30 69
Ole Miss 0.9354 25 0.8590 22 1.29 70
Boston College 0.8822 70 0.4971 67 1.18 71
South Florida 0.8914 58 0.5140 64 1.02 72
Maryland 0.9443 23 0.7120 36 0.72 73
Georgia Tech 0.9061 48 0.5873 52 0.48 74
Arizona 0.8931 55 0.6528 43 0.35 75
Southern Miss 0.8654 82 0.3091 84 0.25 76
Wake Forest 0.8866 64 0.5244 61 -0.09 77
Wyoming 0.8302 123 0.1587 119 -0.11 78
Nevada 0.8618 86 0.2482 93 -0.63 79
Northwestern 0.9055 49 0.6078 49 -1.11 80
Florida State 1.0001 7 0.9729 5 -1.43 81
Indiana 0.8926 56 0.5759 54 -1.47 82
Army 0.8156 128 0.1465 121 -1.61 83
UL-Lafayette 0.8379 114 0.2117 101 -1.64 84
Baylor 0.9145 39 0.7713 31 -1.87 85
Florida International 0.8834 68 0.2892 87 -1.92 86
Tennessee 0.9806 11 0.9117 14 -2.19 87
Colorado 0.8940 54 0.5432 58 -2.42 88
Air Force 0.7863 130 0.1457 122 -2.74 89
Tulane 0.8698 79 0.2960 86 -3.26 90
UCLA 0.9709 14 0.9020 16 -3.55 91
Kansas State 0.8884 61 0.5227 62 -4.08 92
Arkansas 0.9306 30 0.7769 29 -4.35 93
SMU 0.8590 88 0.3617 77 -4.44 94
Louisiana Tech 0.8740 77 0.3268 81 -4.48 95
North Carolina 0.9228 33 0.7912 27 -4.62 96
Western Michigan 0.8717 78 0.4070 74 -6.28 97
UL-Monroe 0.8378 116 0.1650 118 -7.14 98
Old Dominion 0.8501 100 0.2283 95 -9.15 99
Hawaii 0.8482 102 0.1688 117 -9.32 100
Tulsa 0.8519 96 0.2663 90 -9.62 101
Navy 0.8240 126 0.1964 109 -9.81 102
Illinois 0.8882 62 0.5336 60 -10.59 103
New Mexico 0.8338 120 0.1911 112 -11.84 104
Texas State 0.8411 109 0.2209 98 -12.03 105
Western Kentucky 0.8571 91 0.2682 89 -12.25 106
East Carolina 0.8588 89 0.3431 79 -12.78 107
Kansas 0.8813 71 0.4451 71 -12.95 108
Louisville 0.9256 31 0.7218 35 -12.99 109
UNLV 0.8419 108 0.2116 102 -13.13 110
Colorado State 0.8637 83 0.3220 82 -13.21 111
Charlotte 0.8306 122 0.1438 124 -13.57 112
Massachusetts 0.8361 117 0.2093 104 -13.84 113
Coastal Carolina 0.8274 124 0.0870 130 -14.16 114
South Alabama 0.8622 85 0.2017 106 -15.04 115
Rutgers 0.8808 73 0.5358 59 -15.60 116
Ball State 0.8339 119 0.1933 111 -15.77 117
Liberty 0.8029 129 0.1173 126 -16.33 118
Akron 0.8382 113 0.1090 128 -16.60 119
Central Michigan 0.8394 111 0.1823 115 -17.49 120
Kent State 0.8439 107 0.1327 125 -17.57 121
Georgia State 0.8406 110 0.1728 116 -17.75 122
Oregon State 0.8864 65 0.5473 57 -18.12 123
New Mexico State 0.8267 125 0.1089 129 -18.15 124
San Jose State 0.8544 94 0.3126 83 -18.52 125
UTSA 0.8513 97 0.2471 94 -19.00 126
Bowling Green 0.8538 95 0.2272 96 -19.76 127
UTEP 0.8218 127 0.1103 127 -21.52 128
Rice 0.8378 115 0.1839 114 -22.17 129
Connecticut 0.8456 105 0.2201 99 -25.91 130

13.

Not a massive difference, by any means. Using this method would have still resulted in S&P+ projections missing wildly on USC and FSU. There are always some drastic underachievers, obviously. But if it makes the projections better, it makes the projections better, so I’m going to play with this some more.

14.

That Pareto should have charged a monthly subscriber fee.