True Fantasy Points: 2015 Running Backs

A stats-based guide for identifying "true" running back talent in fantasy drafts.

Last week, I detailed how my true fantasy points (TFP) system applies during fantasy draft season as much as it does during the NFL regular season. In short, we can compare the "true" fantasy-relevant stats each player has accumulated with his current team to the stats that David Dodds projects in 2015, and thereby identify players that Dodds might have ranked too high or too low. Today, I'm going to pivot from quarterback projections to running back projections.

The backbone of TFP is the set of "stabilization points" (read: trustworthiness thresholds) for fantasy-relevant stats. For running backs, they are as follows:

Even without doing the sophisticated math I'll be presenting shortly, this information is still quite useful. And here's why: It tells you how trustworthy a player's past performance is for predicting his future performance; and that's kind of important this time of year. As an example, let's consider Adrian Peterson. He's run the ball 2,057 times and run 1,634 routes with the Vikings. With respect to the running game, Peterson's carry total barely exceeds the YPC stabilization point, but far exceeds the RuTD% stabilization point. Therefore, we can conclude that his 4.18 RuTD% through 2014 is far more trustworthy than his 4.96 YPC. With respect to receiving, Peterson's routes run total exceeds all three of the above thresholds, but to varying degrees, which means that his 13.0% RPRR in Minnesota is slightly more trustworthy than his 1.08 YPRR, and both are far more trustworthy than his 0.39% TDPRR.

Furthermore, we can also use this information -- again, without sophisticated math -- to compare our confidence in Peterson's overall rushing projection with that of another running back. For instance, C.J. Anderson's only had 186 carries and 231 routes run so far in a Broncos uniform. Because both of those totals are far below Peterson's totals in Minnesota, we can trust Peterson's past performance far more than we can trust Anderson's. That doesn't mean Peterson's going to be better than Anderson in 2015. It just means that, Peterson's more likely than Anderson to finish 2015 -- barring injury -- with per-attempt and per-route stats closer to what he's produced in the past.

That was the simple version. The real fun starts when we turn up the math volume; not all the way to 11 mind you, but from, say, 1 to 2. At that level, we can calculate each running back's "true" fantasy-relevant performance in the past, and then use Dodds' projections for attempts and routes run to calculate each running back's TFP projection for 2015.

I know. You're thinking, "Wait a second. Dodds projects receptions, not routes run!" You're absolutely right; more on that later.

true Rush Td%, rprr, yprr, & tprr for 2015

In the meantime, let's start with the first part of the TFP equation. Below is a table (sorted by "True YPC") showing "true" performance for the 34 running backs who are a) on the same team as last season, and b) either a starter or among the top two members of a projected running-back-by-committee:

NameTmPrev AttPrev RuYdsPrev RuTDsPrev RRPrev RecsPrev RecYdsPrev RecTDsTrue YPCTrue RuTD%True RPRRTrue YPRRTrue TDPRR
Jamaal Charles KC 1245 6840 38 1723 263 2267 19 4.88 3.10% 15.3% 1.31 0.88%
Adrian Peterson MIN 2057 10196 86 1634 208 1700 5 4.73 3.94% 13.0% 1.08 0.39%
Justin Forsett BAL 235 1266 8 304 44 263 0 4.60 3.25% 14.7% 1.07 0.41%
Jeremy Hill CIN 222 1124 9 176 27 215 0 4.56 3.41% 15.3% 1.26 0.45%
Jonathan Stewart CAR 1041 4827 30 974 130 1084 5 4.55 3.00% 13.6% 1.15 0.52%
LeGarrette Blount NE 213 1053 10 96 6 56 0 4.54 3.56% 11.9% 1.11 0.49%
Latavius Murray OAK 82 424 2 133 17 143 0 4.53 3.11% 14.1% 1.22 0.47%
Lamar Miller MIA 443 2051 11 620 70 490 1 4.52 2.91% 12.1% 0.95 0.39%
C.J. Anderson DEN 186 887 8 231 34 324 2 4.52 3.43% 14.9% 1.33 0.59%
Joseph Randle DAL 104 504 5 68 12 84 0 4.52 3.41% 15.9% 1.27 0.50%
Arian Foster HOU 1390 6294 53 1478 227 2047 12 4.51 3.61% 15.3% 1.37 0.69%
Alfred Morris WAS 876 3955 28 568 37 310 0 4.51 3.19% 8.5% 0.80 0.35%
Tre Mason STL 179 765 4 121 16 148 1 4.48 2.99% 14.4% 1.26 0.56%
Chris Ivory NYJ 376 1648 9 220 20 133 1 4.48 2.90% 11.7% 1.00 0.52%
Carlos Hyde SF 83 333 4 113 12 68 0 4.48 3.37% 13.4% 1.09 0.48%
Devonta Freeman ATL 65 248 1 125 30 225 1 4.48 3.05% 18.9% 1.43 0.56%
Montee Ball DEN 174 732 5 213 29 207 0 4.48 3.13% 14.3% 1.15 0.44%
Eddie Lacy GB 530 2317 20 551 77 684 4 4.47 3.45% 14.3% 1.25 0.60%
Isaiah Crowell CLE 148 607 8 156 9 87 0 4.47 3.59% 10.7% 1.03 0.46%
Fred Jackson BUF 1281 5642 30 1748 321 2637 7 4.46 2.63% 18.1% 1.47 0.45%
Marshawn Lynch SEA 1346 5930 54 1110 145 1229 8 4.46 3.74% 13.3% 1.14 0.63%
Danny Woodhead SD 121 463 2 340 81 643 6 4.46 2.96% 21.0% 1.60 0.83%
Andre Ellington ARI 318 1312 6 460 85 766 3 4.45 2.77% 17.6% 1.51 0.57%
Rashad Jennings NYG 167 639 4 172 30 226 0 4.45 3.03% 16.3% 1.29 0.46%
Bishop Sankey TEN 153 572 2 146 18 133 0 4.45 2.84% 13.9% 1.16 0.47%
Charles Sims TB 66 185 1 116 19 190 0 4.45 3.04% 15.7% 1.37 0.48%
Giovani Bernard CIN 338 1375 10 521 99 863 5 4.44 3.11% 18.1% 1.52 0.67%
LeVeon Bell PIT 534 2221 16 820 128 1259 3 4.43 3.11% 15.5% 1.46 0.46%
Mark Ingram NO 582 2426 20 364 53 288 0 4.43 3.31% 14.8% 1.01 0.40%
Joique Bell DET 470 1923 17 803 139 1354 1 4.42 3.37% 16.9% 1.57 0.36%
Doug Martin TB 581 2404 14 561 74 602 1 4.42 2.83% 13.7% 1.14 0.41%
Knile Davis KC 204 705 10 196 27 222 1 4.40 3.59% 14.4% 1.22 0.53%
Alfred Blue HOU 169 528 2 109 15 113 1 4.39 2.79% 14.6% 1.21 0.57%
Matt Forte CHI 1815 7699 41 2453 445 3722 16 4.38 2.51% 18.0% 1.49 0.62%

So we're on the same page, let's use Lamar Miller as an example of how to read the table. In Miller's three-year Dolphins career, he's run the ball 443 times for 2,051 yards, which translates to 4.63 YPC. However, if we add 1,978 carries' worth of 4.50 YPC (i.e., the same-team league-average over the stabilization-point number of carries), it turns out his True YPC is 4.52, or 0.11 YPC lower than his actual performance thus far in Miami. Apply this same math adjustment to Miller's total rushing touchdowns, receptions, receiving yards, and receiving touchdowns over the past three seasons, and you end up with his "true" performance rates on the right side of the table.

The running back that immediately jumped out at me when I first looked at this table was Matt Forte. Seems to me that his "true" rushing stats lay bare the reality that no one should be -- or should have been -- drafting him for his prowess as a runner. Although YPC is the least trustworthy rushing stat of all, Forte's one of only two running backs in the table with a carry total near YPC's trustworthiness threshold. And yet his YPC is 0.35 lower than, and on the other end of the spectrum from, that other running back: Adrian Peterson.

As was the case with my earlier comparison with C.J. Anderson, however, none of this means Peterson will score more fantasy points than Forte this season; just that we can be confident Peterson is the better runner. Now, to apply all of the above to fantasy projections, we need to turn that math volume up to 3 and apply these "true" rates to Dodds' projections for rushing and receiving "attempts."

true fantasy projections for 2015

Why did I just put attempts in quotation marks? Well, it relates back to the fact that Dodds doesn't project routes run. Luckily, all is not lost. There's a mathematical way around this problem if we turn the dial up to 4.

The solution is to accept Dodds' receptions projection -- not to be confused with Matt Harmon's Reception Perception -- as true, and then reverse-engineer each running back's projected number of routes run. The easiest way to explain what I mean is by way of an example. And since I just slagged off Forte's "true" rushing performance, let's consider his "true" receiving performance. The calculation goes a little something like this:

1. Dodds projects 62 receptions.
2. Per the above table, Forte's True RPRR is 18.0%.
3. If Forte actually ends up with 62 receptions in 2015, then his True RPRR means he ran 345 routes (i.e., 62 divided by 0.18).
4. If he actually runs 345 routes, then his True 1.49 YPRR translates to 515 "true" receiving yards this season.
5. If he actually runs 345 routes, then his True 0.62% TDPRR translates to 2 "true" receiving touchdowns.
It's not ideal, but I think this bit of mathematical reasoning holds up.2

Of course, the byproduct of using said logic is that my "true" receptions projection and Dodds' receptions projection are equal by definition. And since receptions is the only stat that separates standard Footballguys (FBG) scoring from PPR scoring, the discrepancies between our running back projections are -- again, by definition -- the same for both scoring systems. Given that, on average, only about 20 percent of PPR scoring comes from receptions, I'm not worried in the slightest that the results of this exercise are invalid.

Speaking of which, here's a table of said results (based on standard FBG scoring and sorted by the "Diff" column):

NameTmTrue RuYdsTrue RuTDsTrue ReYdsTrue ReTDsDodds RuYdsDodds RuTDsDodds ReYdsDodds ReTDsFBGRkTFPRkDiffRk
Jeremy Hill CIN 1094 8 206 1 1176 10 200 1 203.6 8 183.5 10 +20.1 1
LeVeon Bell PIT 1019 7 547 2 1024 10 493 3 229.7 2 209.7 5 +20.0 2
Jamaal Charles KC 1148 7 403 3 1210 9 353 3 228.3 3 215.2 3 +13.1 3
Marshawn Lynch SEA 1160 10 300 2 1170 11 308 2 225.8 4 214.3 4 +11.5 4
Lamar Miller MIA 950 6 267 1 945 7 265 2 175.0 13 165.0 13 +10.0 5
Joseph Randle DAL 1017 8 208 1 1046 9 169 1 181.5 10 173.4 11 +8.1 6
Eddie Lacy GB 1186 9 343 2 1179 10 343 2 224.2 5 217.5 2 +6.7 7
Mark Ingram NO 1040 8 172 1 1022 10 155 0 177.7 11 171.8 12 +5.9 8
Latavius Murray OAK 951 7 189 1 977 7 176 1 163.3 14 157.6 17 +5.7 9
C.J. Anderson DEN 1086 8 366 2 1068 9 349 2 207.7 6 204.3 7 +3.4 10
Adrian Peterson MIN 1278 11 291 1 1215 12 249 2 230.4 1 227.1 1 +3.3 11
Tre Mason STL 515 3 176 1 489 4 176 1 96.5 28 94.4 31 +2.1 12
Matt Forte CHI 1051 6 515 2 1044 7 490 2 207.4 7 205.4 6 +2.0 13
Justin Forsett BAL 919 6 451 2 940 6 459 2 187.9 9 186.4 9 +1.5 14
Alfred Blue HOU 571 4 133 1 520 4 128 1 94.8 29 95.8 30 -1.0 15
Rashad Jennings NYG 823 6 198 1 749 6 190 1 135.9 21 139.9 22 -4.0 16
Joique Bell DET 730 6 251 1 660 6 227 1 130.7 24 134.8 25 -4.1 17
Danny Woodhead SD 335 2 397 2 293 2 406 2 93.9 30 98.8 29 -4.9 18
LeGarrette Blount NE 954 7 65 0 903 8 46 0 142.9 18 148.5 19 -5.6 19
Knile Davis KC 484 4 110 0 451 4 111 0 80.2 34 86.0 34 -5.8 20
Fred Jackson BUF 424 3 269 1 352 2 238 2 83.0 33 89.2 33 -6.2 21
Giovani Bernard CIN 666 5 361 2 630 4 348 2 133.8 22 140.2 21 -6.4 22
Chris Ivory NYJ 896 6 170 1 840 6 142 1 140.2 19 146.8 20 -6.6 23
Andre Ellington ARI 823 5 377 1 731 5 374 2 152.5 16 159.3 16 -6.8 24
Carlos Hyde SF 941 7 189 1 903 7 150 1 153.3 15 160.4 14 -7.1 25
Arian Foster HOU 745 6 223 1 743 5 210 1 131.3 23 139.3 23 -8.0 26
Devonta Freeman ATL 717 5 302 1 624 4 304 2 128.8 25 138.2 24 -9.4 27
Montee Ball DEN 560 4 113 0 488 4 106 0 83.4 32 93.3 32 -9.9 28
Charles Sims TB 534 4 237 1 462 3 227 1 92.9 31 103.9 28 -11.0 29
Alfred Morris WAS 1171 8 179 1 1079 8 158 1 177.7 11 189.5 8 -11.8 30
Doug Martin TB 840 5 151 1 732 6 130 0 122.2 26 134.5 26 -12.3 31
Isaiah Crowell CLE 895 7 146 1 820 7 123 0 136.3 20 151.1 18 -14.8 32
Jonathan Stewart CAR 955 6 212 1 903 5 183 1 144.6 17 160.3 15 -15.7 33
Bishop Sankey TEN 711 5 192 1 608 4 170 0 101.8 27 122.3 27 -20.5 34

Again making sure we're on the same page, let's use Jeremy Hill as an example of how to read the table. If we apply Hill's "true" stats from the earlier table to Dodds' projections for rushing attempts and implied routes run, he has a TFP projection of 183.5 based on a rushing stat line of 240-1,094-8 and a receiving stat line of 25-206-1. Meanwhile, Dodds projects 203.6 points for Hill based on a rushing stat line of 240-1,176-10 and a receiving stat line of 25-200-1. That means Dodds projects 20.1 more points than does TFP, and that's the biggest discrepancy among running backs in 2015 who meet my earlier-defined criteria.

Before identifying a couple of over- and under-valued running backs according to the discrepancy between Dodds and TFP, it's worth noting a general aspect of the table that you might have overlooked: At most, our dueling running back projections are about 20 total points away from each other. In a 16-game season, that translates to ± 1.25 points per game -- at the extremes. And to put this in context, the "Diff" column in my articles on quarterback projections topped out at around +40, not around +20. In short (and as usual), Dodds has it mostly right.

That said, I do think there are two running backs who a) reside on the extremes of the table, and b) have had enough opportunities in both the rushing game and the passing game to pique my interest.

jamaal charles

From the first table, we learned that Charles' True YPC is 4.88; Dodds projects him to rush for 5.15 YPC in 2015. If this was, say, C.J. Anderson, I'd be fine with it, but Charles has carried the ball 1,148 times for the Chiefs. Yes, that's about 800 less than the YPC stabilization point, but I remain skeptical that Charles will increase his YPC this year, which is what Dodds' projection suggests.

Then there's Dodds' projection of 9 touchdowns, which translates to a 3.83 RuTD%. Again, as the first table showed, Charles' True RuTD% is 3.10%. That's a veritable quantum leap. And given that Charles' carry total in Kansas City far exceeds the RuTD% stabilization point, that's a leap I'm not inclined to take.

jonathan stewart

Stewart is somewhat of an anti-Charles: Dodds projects him at 4.30 YPC and a 2.38 RuTD%, whereas his True YPC with the Panthers is 4.55 and his True RuTD% is 3.00%. Furthermore, as with Charles, this isn't a situation where we're talking about a relatively green-with-the-team running back like C.J. Anderson. Given his 1,000+ carries in Carolina, Stewart's True RuTD% is beyond the threshold for trustworthiness going forward.

1 For those who are new to TFP, routes run is a stat charted by Pro Football Focus that counts the number of snaps in which a player -- regardless of his position -- ran a route. The reason I use RR as the fundamental measure of "receiving attempts" is because Chase Stuart and I have found that per-route stats are far more consistent over time than per-target stats. And if we're going to try to project future performance from past performance, the more consistent a performance metric is over time, the more accurate our projections will be.

2 The explanation here is complicated, and has to do with the fact that YPRR and TDPRR don't have the same stabilization point. If you'd like further (esoteric) explanation, feel free to contact me.