True Fantasy Points: 2018 Running Back Projections - Footballguys

A stats-based guide for separating skill from luck in running back projections.

This is the second installment in the True Fantasy Points (TFP) series for 2018. If you missed the previous article on quarterbacks, click here. And if you want to learn more about the statistical methodology underlying the series, click here or here.

As was the case with quarterbacks on Monday, today's mission is to use TFP to identify running backs whose average draft positions (ADP) are out of whack with their "true" value. As you'll see, 2018 provides more fodder for the overvalued side of the spectrum than the undervalued side. Furthermore, you'll come to learn that the randomness of running back stats results in David Dodds' projections being far more in line with TFP than they were for quarterbacks.

Below is a table showing Dodds' projections, TFP, and their difference for 44 running backs who meet the following criteria:

  • They're on the same team as they were last season.
  • They're projected for at least 200 "opportunities," which is defined as rushing attempts plus routes run.
Dodds' Stats True Stats Points
Running Back Team RuYds RuTDs ReYds ReTDs RuYds RuTDs ReYds ReTDs Dodds TFP Diff
Alvin Kamara NO 822 7 653 5 684 5 647 3 217.7 182.0 +35.7
Todd Gurley LAR 1296 11 611 4 1221 9 608 2 281.9 252.0 +29.9
Le'Veon Bell PIT 1142 10 608 3 1142 7 654 2 254.8 237.4 +17.4
Kareem Hunt KC 1104 8 403 2 1044 7 418 2 212.5 197.4 +15.1
Mark Ingram II NO 666 5 248 2 637 5 236 1 133.4 121.8 +11.6
Dalvin Cook MIN 1088 8 300 2 1080 7 298 1 198.8 188.7 +10.1
Devonta Freeman ATL 920 8 383 2 918 7 356 2 190.3 181.6 +8.7
Christian McCaffrey CAR 758 5 570 4 811 5 520 3 189.2 181.7 +7.5
T.J. Yeldon JAX 372 2 195 1 342 2 186 1 75.9 68.5 +7.4
Bilal Powell NYJ 466 4 205 1 473 3 206 1 94.7 88.2 +6.5
Duke Johnson Jr CLE 229 1 519 3 236 1 498 2 98.8 92.3 +6.5
Rex Burkhead NE 514 5 292 2 537 4 276 1 119.6 113.8 +5.8
Matt Breida SF 423 4 130 1 429 3 129 1 81.7 76.1 +5.6
Giovani Bernard CIN 389 3 352 2 398 2 357 1 103.5 98.4 +5.1
Ty Montgomery GB 287 3 320 2 305 2 317 1 87.1 83.2 +3.9
Leonard Fournette JAX 1195 10 420 3 1260 9 391 2 233.5 230.0 +3.5
Melvin Gordon III LAC 1076 8 451 3 1148 7 420 2 217.5 214.3 +3.2
Austin Ekeler LAC 339 2 261 2 342 2 248 1 83.4 80.6 +2.8
Marshawn Lynch OAK 657 7 131 1 683 5 161 1 122.0 119.6 +2.4
Chris Thompson WAS 215 1 468 3 218 1 466 3 95.3 93.3 +2.0
Tarik Cohen CHI 419 2 413 2 422 3 392 2 108.4 107.5 +0.9
Alex Collins BAL 978 8 218 1 1003 7 232 1 170.6 169.9 +0.7
Aaron Jones GB 422 4 105 1 433 3 109 1 77.3 76.6 +0.7
Kenyan Drake MIA 834 5 296 2 825 6 298 1 155.0 154.6 +0.4
Chris Carson SEA 548 4 224 2 581 4 232 1 110.8 110.7 +0.1
Joe Mixon CIN 878 7 264 1 926 6 275 1 162.2 163.2 -1.0
Tevin Coleman ATL 580 5 323 2 595 5 358 2 132.3 133.5 -1.2
Jordan Howard CHI 988 9 170 1 1061 7 211 1 171.6 173.6 -2.0
D'Onta Foreman HOU 470 3 151 1 486 3 152 1 84.9 87.1 -2.2
Derrick Henry TEN 959 8 205 1 999 7 226 1 170.4 172.8 -2.4
Javorius Allen BAL 332 2 228 1 359 2 223 1 75.8 78.2 -2.4
James White NE 388 3 336 2 423 3 355 2 104.2 106.8 -2.6
Ezekiel Elliott DAL 1376 12 446 3 1403 11 494 2 269.2 272.0 -2.8
Theo Riddick DET 205 1 353 2 232 2 339 2 73.8 77.7 -3.9
Devontae Booker DEN 516 3 282 1 548 4 294 1 106.8 112.3 -5.5
LeSean McCoy BUF 725 5 207 1 770 5 210 1 129.8 135.3 -5.5
Corey Clement PHI 453 3 215 1 469 3 242 1 94.4 100.1 -5.7
Jay Ajayi PHI 884 6 178 1 907 6 198 1 147.6 153.9 -6.3
Jamaal Williams GB 618 5 254 1 662 5 270 1 122.6 129.1 -6.5
Marlon Mack IND 672 5 279 2 729 5 278 1 131.1 137.9 -6.8
Peyton Barber TB 758 6 218 1 813 5 234 1 135.4 143.3 -7.9
Lamar Miller HOU 796 5 295 2 851 5 305 2 147.5 155.6 -8.1
Latavius Murray MIN 520 4 158 1 554 4 177 1 93.0 101.4 -8.4
David Johnson ARI 1054 9 641 3 1093 10 760 4 240.9 266.8 -25.9

As an example of how to read this table, focus on the row associated with Joe Mixon. Dodds projects 1,142 total yards and 8 touchdowns, whereas the regression-to-the-mean-based TFP system projects 1,201 total yards and 7 touchdowns. Applying Footballguys' standard scoring system to these totals means Dodds' projects 162.2 points and TFP projects 163.2 points for Mixon. As these are only 1.0 points apart, one can arrive at two statistical conclusions:

  1. Dodds has optimally accounted for the randomness of running back stats and properly regressed them to the mean.
  2. Mixon is neither overvalued or undervalued in fantasy drafts. He's just right, likely being selected nearly exactly where he should be.

With everyone on the same page, let's discuss the anti-Mixons, i.e., running backs who aren't valued correctly right now.

The Four Horsemen of 2018 Regression to the Mean

Over the course of the offseason and into August, Todd Gurley, Le'Veon Bell, Ezekiel Elliott, and David Johnson have comprised a quartet of running backs coming off the board in the first four picks overall. The thing is, these four aren't created equally. To wit (emphasis added):

  • Gurley -- 4.2 True Yards-per-Carry (YPC), 3.2% True Touchdowns-per-Carry (TDPC), 14.7% True Receptions per Route Run (RPRR), 1.4 True Yards per Route Run (YPRR), 0.5% True Touchdowns per Route Run (TDPRR)
  • Bell -- 4.3 True YPC, 2.8% True TDPC, 16.8% True RPRR, 1.4 True YPRR, 0.5% True TDPRR
  • Elliott -- 4.4 True YPC, 3.5% True TDPC, 12.1% True RPRR, 1.2 True YPRR, 0.6% True TDPRR
  • Johnson -- 4.3 True YPC, 4.0% True TDPC, 16.1% True RPRR, 1.7 True YPRR, 0.6% True TDPRR

Based on the above, it seems the rankings are almost exactly backward. The combination of Elliott's projected workload and True YPC, both of which rank first among these four running backs, suggests he should be getting selected first or second, yet his average draft position (ADP) is currently third. Similarly, Johnson, who the table shows as more undervalued than any other running back, has the truest "skill" at gaining receiving yards and scoring touchdowns. Dodds' projection of 1.4 YPRR is well below Johnson's "true" rate, as is his 3.4% TDPC projection. Johnson should also be getting selected first or second, and yet his current ADP is fourth.

Meanwhile, Gurley and Bell are slated for among the most regression to the mean of any running backs according to TFP, and yet they're average draft positions are first and second, respectively. In the case of Gurley, his 4.7% TDPC in 2017 should be regressed nearly twice as much as Dodds' 3.9% projection, i.e., another 0.7% downward. Ditto in the case of Bell, whose Dodds-implied 3.8% TDPC is wildly out of step with his 2.8% True TDPC. (Note: It's not often that Dodds and TFP match exactly with respect to yardage, so want to point out this statistical anomaly that may only interest me.)

At this point, the article in past years has pivoted the discussion towards two more undervalued running backs and two more overvalued running backs. But, as the table shows, no other undervalued running back besides David Johnson even reaches double digits in the "Diff" column. Therefore, the remaining discussion will focus exclusively on overvalued running backs.

Overvalued: Alvin Kamara

Kamara is the perfect teaching example of regression to the mean in YPC. It's an underappreciated fact that it takes nearly 2,000 carries for YPC to represent a 50/50 split between "true" skill and randomness. This renders Kamara's 6.1 YPC last season almost completely meaningless. Dodds, to his credit, did drop that value to 5.3 YPC for his 2018 projection, but it turns out this isn't nearly enough. Given the league average of 4.3 YPC, Kamara's True YPC is actually only 4.4. One can see the effect of this mathematical difference in the table, where 155 carries produce 822 rushing yards according to Dodds, whereas it produces only 684 rushing yards according to TFP.

Overvalued: Kareem Hunt

Dodds' projections for Hunt also suffers from an underestimation of YPC randomness. He regresses Hunt's 4.9 YPC in 2017 to 4.6 for 2018, but Hunt's True YPC is actually 4.4. And although not as random as rushing yards, rushing touchdowns is also a noisy stat. That's why Hunt's 2.9% TDPC should be regressed to his 2.8% True TDPC rather than increased to 3.4% as Dodds suggests.

Overvalued: Mark Ingram II

YPC is also implicated in Ingram's difference between Dodds' projection and TFP, though not as much as was the case for Kamara and Hunt. Instead, what hurts Ingram with respect to TFP is his 0.5% True TDPRR, which is almost half as large as Dodds' projected 0.8% TDPRR.

Overvalued: Dalvin Cook

TDPC is why Cook's TFP is 10.1 points below Dodds' projection. Cook's True TDPC is 2.9%, whereas Dodds's projection implies a 3.2% TDPC.