Measuring and Ranking the Running Back and Wide Receiver Draft Class of 2015

Every year in February, a group of NFL front office personnel, scouting organizations, media members, and the nation's top collegiate athletes get together for the NFL Combine. In this weeklong event, players hoping to make the leap to the NFL are measured, analyzed, and scrutinized by the movers and shakers of the league. Yet, many have wondered whether these movement tests performed in sterile conditions -- in shorts and t-shirts, free of helmets and pads -- have any validity in predicting future NFL performance.

With a large coffee in one hand and the combine results for all 1,091 wide receivers and running backs from the past 16 seasons (courtesy of mockdraftable.com) on my computer in front of me, I decided to tackle the task of assimilating the vast metrics measured at the combine for each player and converting this information into an actionable predictor of NFL potential.

The Math

When we glance at combine numbers, our intuition gives us a general sense of a "good" combine performance versus a "bad" one. But these are subjective measures. Can we define these in a quantitative manner? First, I asked, "What exactly constitutes a 'good' combine performance? What separates the 'good' from the 'great', and the 'great' from the 'elite'?"

I decided to define the quality of an athlete's combine performance as the fold increase above what would be expected for an NFL hopeful of that athlete's size to produce. I chose to normalize everything to size to take into account a condition that makes intuitive sense: if two athlete's both run a 4.4 40-yard dash, but one athlete weighs 265 pounds and the other weights 190 pounds, we would consider the performance by the 265 pound athlete much more impressive. Therefore, I thought that rather than compare performances at face value, our analysis should somehow factor in player size.

So from the large set of historical data at my disposal, I was able to derive a set of equations to determine an athlete's expected combine numbers based on that athlete's size (specifically, weight), which served as the baseline from which I then assessed that player's actual combine performance:

Estimated 40 = (.0022 x Weight) + 4.0726
Estimated Bench = (.1313 x Weight) - 9.825
Estimated Vertical = (-.0102 x Weight) + 37.264
Estimated Broad = (-.0244 x Weight) + 124.08
Estimated 3-Cone = (.004 x Weight) + 6.1036
Estimated 20ss = (.0026 x Weight) + 3.688

So a fair question to ask would be: "Just how valid are these equations in predicting combine performance?" To test this, I took the average weight of all 1,091 running backs and wide receivers (206.6 pounds) and calculated what the combine performances would be for players of this size. I then compared these numbers to the actual average of the combine performances for all players from the last 16 seasons weighing either 205, 206, or 207 pounds (n = 63 players). The results are listed below:

	Weight	40-yd	Bench	Vert	Broad	3Cone	20ss
Predicted	206.6	4.527	17.301	35.157	119.039	6.957	4.225
Actual (Avg.)	205.9	4.534	18.861	35.402	118.982	7.018	4.215

From the data above, we see that these equations do an excellent job of predicting combine performance for a given player's size. Using these equations, I was then able to assess how well a player's actual performance differed from their predicted performance. Players performing better than their expected combine numbers in a given event received a positive score for that event (weighted according to the dynamic range of the particular skill being measured), and conversely, players performing worse than their expected combine numbers for a given event received a negative weighted score for that event. Specifically, the scores for each player for any particular event was obtained using the following formulas:

40 = -(Observed - Expected)/(Range)
Bench = (Observed - Expected)/(Range)
Vertical = (Observed - Expected)/(Range)
Broad = (Observed - Expected)/(Range)
3-Cone = -(Observed - Expected)/(Range)
20ss = -(Observed - Expected)/(Range)

Range is defined as the upper limit of all combine performances for all wide receivers and running backs subtracted by the lower limit for a given event.

Wide Receiver and Running Back Ranks According to PACE

Taking the average scores across all events, I then obtained what I call my Performance Above Combine Expectations (PACE) score. This allowed me to rank all 1,091 wide receivers and running backs from 1999-2015 on the basis of their combine performances, with some of the notable names at the top five percent of all-time PACE scores listed below:

All-Time Top 50	Â	Â	Â	Â
Year	Name	Position	PACE Score	All-time Rank
2007	Calvin Johnson	WR	0.432	1
2001	Chris Chambers	WR	0.355	2
2003	Justin Fargas	RB	0.3428	3
1999	Edgerrin James	RB	0.288	6
2001	Santana Moss	WR	0.275	8
2010	Demaryius Thomas	WR	0.244	11
2014	Jerick McKinnon	RB	0.243	12
2013	Christine Michael	RB	0.221	14
2010	Ben Tate	RB	0.213	16
2008	Chris Johnson	RB	0.212	17
2015	Jaelen Strong	WR	0.206	20
2004	Kevin Jones	RB	0.205	24
2015	Chris Conley	WR	0.200	27
2015	Ameer Abdullah	RB	0.196	29
2011	Julio Jones	WR	0.187	37
2010	C.J. Spiller	RB	0.186	38
2008	Donnie Avery	WR	0.186	39
2015	Sammie Coates	WR	0.185	40
2003	Andre Johnson	WR	0.179	44
2012	Ryan Broyles	WR	0.175	46
2013	Cordarrelle Patterson	WR	0.174	48

Not surprisingly, Calvin Johnson was the top performer according to the PACE metric, achieving a score of 0.43, a figure nearly 25% greater than the second-best performer, wide receiver Chris Chambers.

Some other notable names at the top of this list include athletic freaks who have emerged recently as stars in the NFL, including Demaryius Thomas, Julio Jones, and C.J. Spiller, veterans that have shown long-term productivity in the league including the Colts franchise leader in rushing yards, Edgerrin James, Pro Bowl wideouts Santana Moss and Andre Johnson, and 2014 breakouts including Jerick McKinnon.

Predictive Value of the PACE Metric

To further test the validity of the PACE metric as a predictor of on-field success, I decided to do a case study of the 2011 rookie class. I chose this class because we have four seasons worth of statistics for this group of players that we can then use to assess the relationship between NFL productivity and combine metrics. I ranked every wide receiver and running back participating in at least four events of the 2011 combine by PACE score, and assessed their productivity for the past four seasons using Total Net Expected Points (NEP).

Below is the list of the top 25 players of the class of 2011 as ranked by the PACE metric, along with their total NEP for each season from 2011-2014, their average NEP per season, and their positional rank (percentile) according to NEP per season:

                                                   

Year	Name	Pos	PACE Score	'11 NEP	'12 NEP	'13 NEP	'14 NEP	Avg NEP	RB Rk	WR Rk
2011	Greg Little	WR	0.1905	55.92	54.72	30.49	8.16	37.32	-	77.8%
2011	Julio Jones	WR	0.1875	81.10	105.55	39.34	141.72	91.93	-	94.4%
2011	Torrey Smith	WR	0.1593	75.36	76.45	97.73	80.96	82.63	-	91.7%
2011	Jordan Todman	RB	0.1563	0.00	-2.02	1.90	6.89	1.69	72.4%	-
2011	Aldrick Robinson	WR	0.1503	0.00	24.16	31.45	0.48	14.02	-	61.1%
2011	Anthony Allen	RB	0.1460	-0.45	2.44	0.00	0.00	0.50	65.5%	-
2011	Edmond Gates	WR	0.1440	2.45	20.76	6.83	0.00	7.51	-	38.9%
2011	DeMarco Murray	RB	0.1305	22.51	3.21	39.45	23.54	22.18	96.6%	-
2011	Delone Carter	RB	0.1190	-25.36	8.46	0.00	0.00	-4.23	24.1%	-
2011	Jon Baldwin	WR	0.1139	21.46	25.17	2.14	0.00	12.19	-	55.6%
2011	Terrance Toliver	WR	0.1056	-	-	-	-	-	-	0.0%
2011	Ryan Whalen	WR	0.1033	1.92	4.89	0.00	0.00	1.70	-	22.2%
2011	Ricardo Lockette	WR	0.0908	8.52	0.00	6.55	12.15	6.81	-	36.1%
2011	Mario Fannin	RB	0.0811	-	-	-	-	-	0.0%	-
2011	Ryan Williams	RB	0.0771	0.00	-16.20	0.00	0.00	-4.05	27.6%	-
2011	Cecil Shorts	WR	0.0769	4.67	85.41	62.43	37.19	47.42	-	83.3%
2011	Kendall Hunter	RB	0.0735	10.78	13.00	-4.13	0.00	4.91	82.8%	-
2011	Niles Paul	WR	0.0735	4.32	10.01	3.33	34.10	12.94	-	58.3%
2011	Greg Salas	WR	0.0692	16.20	0.00	11.48	17.17	11.21	-	52.8%
2011	Austin Pettis	WR	0.0581	17.94	30.61	46.51	13.74	27.20	-	75.0%
2011	Derrick Locke	RB	0.0567	-	-	-	-	-	0.0%	-
2011	Roy Helu	RB	0.0552	4.05	0.65	13.57	25.29	10.89	89.7%	-
2011	A.J. Green	WR	0.0529	100.46	111.17	127.49	78.94	104.52	-	97.2%
2011	Shane Vereen	RB	0.0522	2.40	18.27	37.55	29.61	21.96	93.1%	-
2011	Dane Sanzenbacher	WR	0.0451	27.17	1.01	7.21	6.01	10.35	-	50.0%

10 out of the top 25 players as ranked by PACE score had four-year NEP averages that placed them in the top 25th percentile of the 2011 class, including Pro Bowlers Julio Jones, A.J. Green, and DeMarco Murray, along with established starters such as Torrey Smith, Shane Vereen, and Cecil Shorts.

Conversely, 16 out of the bottom 25 players as ranked by PACE score had four-year NEP averages that placed them in the bottom half of the 2011 class:

                                                   

Year	Name	Pos	PACE Score	'11 NEP	'12 NEP	'13 NEP	'14 NEP	Avg NEP	RB Rk	WR Rk
2011	Owen Spencer	WR	0.0013	-	-	-	-	-	-	0.0%
2011	Graig Cooper	RB	-0.0074	-	-	-	-	-	0.0%	-
2011	Stephen Burton	WR	-0.0111	3.36	3.87	3.73	0.00	2.74	-	25.0%
2011	Stevan Ridley	RB	-0.0130	2.44	13.11	-13.28	-1.38	0.22	58.6%	-
2011	Titus Young	WR	-0.0150	56.55	35.31	0.00	0.00	22.96	-	69.4%
2011	Darren Evans	RB	-0.0355	-	-	-	-	-	0.0%	-
2011	Tori Gurley	WR	-0.0386	0.00	0.00	3.24	0.00	0.81	-	19.4%
2011	Shaun Draughn	RB	-0.0399	0.00	2.59	-1.54	-4.11	-0.76	44.8%	-
2011	Allen Bradford	RB	-0.0446	-1.14	0.00	0.00	0.00	-0.29	51.7%	-
2011	Jerrel Jernigan	WR	-0.0455	0.00	1.80	33.47	0.59	8.97	-	47.2%
2011	Vincent Brown	WR	-0.0484	29.50	0.00	41.05	7.74	19.57	-	63.9%
2011	Vai Taua	RB	-0.0494	-	-	-	-	-	0.0%	-
2011	Evan Royster	RB	-0.0495	18.68	1.50	-2.29	0.00	4.47	79.3%	-
2011	DeAndre Brown	WR	-0.0510	-	-	-	-	-	-	0.0%
2011	Tandon Doss	WR	-0.0511	0.00	10.98	22.50	0.00	8.37	-	41.7%
2011	Armon Binns	WR	-0.0518	0.00	19.82	0.00	0.00	4.96	-	30.6%
2011	Dion Lewis	RB	-0.0545	1.62	5.10	0.00	0.00	1.68	69.0%	-
2011	Johnny White	RB	-0.0557	-4.50	-1.23	0.00	0.00	-1.43	41.4%	-
2011	Dwayne Harris	WR	-0.0632	0.00	17.39	8.88	8.37	8.66	-	44.4%
2011	O.J. Murdock	WR	-0.0714	-	-	-	-	-	-	0.0%
2011	Randall Cobb	WR	-0.0730	31.46	98.04	40.95	118.03	72.12	-	88.9%
2011	Jacquizz Rodgers	RB	-0.1161	3.90	6.47	7.97	3.53	5.47	86.2%	-
2011	Lestar Jean	WR	-0.1302	0.00	14.06	7.85	0.00	5.48	-	33.3%
2011	Mark Ingram	RB	-0.1441	0.27	-6.87	-2.99	7.10	-0.62	48.3%	-
2011	Matt Asiata	RB	-0.1877	0.00	-2.00	-1.84	11.45	1.90	75.9%	-

Only four players, Randall Cobb, Matt Asiata, Jacquizz Rodgers, and Evan Royster, had four-year NEP averages placing them in the top 25th percentile of the 2011 class. To further emphasize the predictive value of the PACE metric to player performance, six of the players from this group actually failed to ever see regular season playing time in the NFL.

A closer look at the data reveals that 17 out of the top 25 players as ranked by the PACE metric ranked in the top half of their class according to their four-year average NEP score. In contrast, only 9 of the bottom 25 players achieved this same feat. For these bottom 25 players, there was instead an enrichment for those ranking in the bottom half of their class according to NEP per season, with 16 of these players clustering into this group:

Percentile Rank (Pos)	Â	Â	Â	Â
Â	100-75th	75-50th	50-25th	25-0
PACE Score Top 25	10	7	3	5
PACE Score Bottom 25	4	5	9	7

In addition, the average NEP scores over the 2011-2014 seasons for the top 25 players as measured by PACE nearly tripled the average NEP scores for the bottom 25 players (23.7 vs. 8.7):

Group	Four Year Avg. of Total NEP
PACE Score Top 25	23.7
PACE Score Bottom 25	8.7

Through this analysis of the 2011 rookie class, we can see at multiple levels that players ranking near the top of the list according to PACE score have a strong tendency to have productive NFL careers, whereas players ranking near the bottom of this list (with a few rare exceptions) have a strong tendency to have either lackluster or even non-existent NFL careers.

Using the PACE Score to Assess the Incoming Class of 2015

So how does the running back and wide receiver class of 2015 stack up when we rank them based on the PACE metric? Below is the list of players clustering into the top 20 (or roughly top 25% of all skill position players) according to PACE score:

Rookie Class of 2015 Top 20	Â	Â	Â	Â
Year	Name	Pos	PACE Score	2015 Rk
2015	Jaelen Strong	WR	0.206064702	1
2015	Chris Conley	WR	0.199988665	2
2015	Ameer Abdullah	RB	0.195970932	3
2015	Sammie Coates	WR	0.184831338	4
2015	David Johnson	RB	0.163048363	5
2015	Tevin Coleman	RB	0.1492875	6
2015	Kevin White	WR	0.124324899	7
2015	Antwan Goodley	WR	0.121578947	8
2015	Kenny Bell	WR	0.110632442	9
2015	Geremy Davis	WR	0.105123048	10
2015	Amari Cooper	WR	0.104148538	11
2015	Rannell Hall	WR	0.10350315	12
2015	Ty Montgomery	WR	0.088294351	13
2015	Tre McBride	WR	0.082919086	14
2015	DeVante Parker	WR	0.08162597	15
2015	Phillip Dorsett	WR	0.081148093	16
2015	Melvin Gordon	RB	0.063459594	17
2015	Jay Ajayi	RB	0.056708064	18
2015	Devin Smith	WR	0.047593305	19
2015	Michael Dyer	RB	0.042384857	20

We see the appearance of some familiar, highly-touted NFL draft prospects on this list including Fred Biletnkoff Award winner Amari Cooper, Doak Walker Award winner Melvin Gordon, and projected first-rounders Kevin White and Jaelen Strong. But where it gets interesting is the presence of names on this list projected to be drafted in the fifth round or later, including Georgia Bulldog wideout Chris Conley and Louisville running back Michael Dyer. These players represent excellent value picks for NFL teams willing to take a chance on them in the upcoming draft.

If history serves as a guide for us, the players on this top 20 list have a high probability of being impact players for the NFL teams that draft them this upcoming April. And, in particular, a few of these aforementioned players -- who have all the athletic skills necessary to succeed at the NFL level -- may represent overlooked candidates with the potential to have careers that vastly outperform their NFL draft position.