Today, I look at running backs drafted since 1984. I use a slightly different way of looking at the data that I think is a little better. I also revisit the QBs and WR/TEs with that method. Instead of considering the number of first-round college teammates that a player has, I consider the total draft value of college teammates at different positions, as determined by Chase’s chart.1 Going this way makes it possible to look at the entire offensive line’s value, for example, rather than just the number of players who were high picks.
For example, according to PFR’s Approximate Value (AV), Ki-Jana Carter is the biggest underachiever at RB relative to his draft position (since 1984). After being drafted #1 in 1995, he generated just nine points of AV in his first five years.2 Carter also had a lot of help from his friends in college. He ranks 10th out of 104 RBs picked in the top 32 in terms of the total value of his college offensive linemen according to my measure. His tight end also went in the top ten in 2005; Carter would be 2nd in total line value if we included TEs. Two of his offensive lineman went in the first round in the following year. Two Penn State fullbacks were drafted that year, too.3 Could Carter have looked better than he was because he ran behind those great college blockers? Or is the NFL success of the running back who ranks fourth in terms of offensive line help (Warrick Dunn) more representative of RBs, in general?
In addition to looking at the offensive line, I’ll consider whether the total value of college teammates at other offensive positions predicts that running backs become overvalued in the draft. While we might think that RBs are particularly dependent on line help, it actually appears that having a great QB is again the one clear predictor for players being overvalued.
As with last week’s post, I use a regression to generate expected AV in the first five seasons according to draft position. I then estimate how a player’s value above expectation (VAE), the difference between actual and expected AV, relates to the quality of his college teammates. The note contains all the details for those interested.4 The big difference today is that I’m looking at all players who were drafted up to pick 224 and all college teammates of that player, rather than focusing just on high picks. For each player, I consider the impact that having better college teammates has on the VAE that a player generates in his first five seasons. I look separately at the total value of a player’s teammates at OL, WR/TE, and at QB.
The table below captures the impact that teammate quality has on how the 545 running backs picked from 1984-2009 performed relative to their draft positions.
|College teammates being worth 20 more points of AV at|
|Predicts a RB's VAE changes by:||-3.06||No clear effect||No clear effect|
In the table, I’ve turned the regression coefficient into the impact of adding 20 points of draft value to the players’ teammates at a given position. What does that mean in English? All else being equal, adding 20 points of draft position value to a running back’s college QB means that the RB should be expected to underperform his draft position by 3.06 points. The 10th pick in the draft has a value of about 20 points. If a running back played with a quarterback who was selected 10th overall rather than say, an undrafted college quarterback, it means that running back will likely be overdrafted and have a VAE of -3.06.
As indicated in the table, the data show no clear effect for a RB’s teammates at OL and WR. I’m not surprised by this for WR, but I am for OL.5 My expectation was that, if there was anything going on, it would appear for OL teammates first and maybe QBs after that. I figured that RBs like Carter and Ron Dayne might have benefited from playing in front of excellent college lines that inflated their college stats. The results seem to suggest that teams figure that stuff out in the draft, not overvaluing RBs who have great college lines. These results do not include TEs with the offensive line, but we get a similar result if we include them.
The individual RBs who had the best college linemen show a similar pattern to the regression estimates. Here are the RBs from the first three rounds who had at least 25 points of value on their offensive lines, as determined by draft position in the same or following year.6 Only players who have at least five years of service are in the regressions, so players such as Trent Richardson and Eddie Lacy (third and first in total line value, respectively) are not included in the table.
Teammate OL Value
|Tyrone Anthony||1984||-10||North Carolina||26.4|
|Lorenzo White||1988||11.6||Michigan St.||40.2|
|Anthony Johnson||1990||-6.9||Notre Dame||27.7|
|Jerome Bettis||1993||19.8||Notre Dame||28.5|
|Reggie Brooks||1993||-4.8||Notre Dame||28.5|
|Ki-Jana Carter||1995||-27.1||Penn St.||29.3|
|Eddie George||1996||36||Ohio St.||34.6|
|Jon Witman||1996||-8.6||Penn St.||29.3|
|Warrick Dunn||1997||29.9||Florida St.||42.5|
|Mike Cloud||1999||-11.1||Boston Col.||26|
|James Jackson||2001||-5.5||Miami (FL)||26.5|
|Clinton Portis||2002||39.4||Miami (FL)||26.5|
13 of the 28 RBs with the best college lines actually performed above expectation in their first five years. Overall, the average performance for these players is very similar to other first-round RBs. Average VAE for these 28 RBs is 1.24, compared to 0.71 for other RBs taken in rounds 1-3. In general, the evidence does not suggest that RBs that drive through truck-sized holes provided by great college lines have excessively lofty status in the draft. [Chase note: I'll add that someone like Montee Ball, who produced absurd statistics in college, was only drafted with the 58th pick in 2012. Anecdotally, it does seem as though NFL teams are well-aware of the benefits college running backs have from playing behind great offensive lines, Alabama running backs aside.]
Things look very different for the impact of a great college QB on the valuation of RBs in the draft. Here are the RBs taken in rounds 1-3 who had a QB with at least 20 points of value according to Chase’s chart.
Teammate QB value
|Ki-Jana Carter||1995||1||-27.1||Penn St.||24.3||Kerry Collins|
|Jerome Bettis||1993||10||19.8||Notre Dame||30.2||Rick Mirer|
|Reggie Brooks||1993||45||-4.8||Notre Dame||30.2||Rick Mirer|
|Alonzo Highsmith||1987||3||-22.6||Miami (FL)||34.6||Vinny Testaverde|
|Cedric Benson||2005||4||-9.9||Texas||27.6||Vince Young|
|Knowshon Moreno||2009||12||4.9||Georgia||34.6||Matthew Stafford|
|Charlie Garner||1994||42||-1.4||Tennessee||28.8||Heath Shuler|
|Joseph Addai||2006||30||29.4||LSU||34.6||JaMarcus Russell|
|Ben Tate||2010||58||-2.5||Auburn||34.6||Cam Newton|
|Rodney Culver||1992||85||-7.3||Notre Dame||30.2||Rick Mirer|
|Tony Brooks||1992||92||-11.6||Notre Dame||30.2||Rick Mirer|
|Jay Graham||1997||64||-11.7||Tennessee||34.6||Peyton Manning|
|Ronnie Harmon||1986||16||0.9||Iowa||24.9||Chuck Long|
|Maurice Morris||2002||54||-4.1||Oregon||27.6||Joey Harrington|
|Keith Jones||1989||62||-3.9||Illinois||34.6||Jeff George|
|Gaston Green||1988||14||-11||UCLA||34.6||Troy Aikman|
|Eric Ball||1989||35||-15.1||UCLA||34.6||Troy Aikman|
On this list, only 4 of the 17 players have positive VAEs. The list also includes the lowest VAE (Carter) and third-lowest VAE (Alonzo Highsmith) for RBs since 1984. Altogether, the mean VAE is -4.59 for these 17 RBs who played in college with the best QBs, compared to 1.14 for the other RBs taken in the first three rounds. The sample size is small, of course, and Highsmith and Carter both dealt with knee injuries. But the 4 out of 17 number is pretty bleak.
Quarterbacks and Receivers
I also checked things out for QBs and WR/TEs, using total value across each position rather than looking at the number of teammates drafted in the first round as in last week’s post. The main results stayed the same when looking at total value. A WR/TE who played in college with a QB with 20 points of additional value in college is overvalued in the draft by about 1.88 points. In other words, going from an undrafted college QB to one picked about 10th predicts a WR/TE does 1.88 points worse in his first five years relative to where the player is selected in the draft.
In contrast, the impact of offensive lines and receivers on quarterbacks continues to be indistinguishable from zero. With the bigger sample that comes from looking at all players picked across seven rounds, the effect size for receivers on QBs also became smaller.7 Overall, the results give no clear evidence that college teammates affect how QBs are valued in the draft.
Overvalued: Backs and Receivers Who Had Excellent College QBs
But there is clear evidence that the reverse is true. In the NFL draft, running backs and wide receivers seem to get overdrafted when they play with really good college QBs. It doesn’t appear to be the case, however, that running behind a great college line leads RBs to be similarly overvalued.
Why might playing with a better college QB cause a RB to be overvalued in the draft, while the line doesn’t have that kind of impact? Allow me to throw out an idea and see what you think. The idea involves some of the more subtle benefits of playing with an excellent QB.
Take Ki-Jana Carter, a RB who may have been overdrafted and who had both great college blockers and a future first-round pick at QB. The big holes opened up by his future first-round teammates at center, left tackle, and tight end were there to be seen on film. It seems like NFL teams have generally made the necessary adjustments to account for the advantage Carter would have had in this situation.
And if that’s right, we might expect teams to account for the more obvious advantages of playing with an excellent QB. Kerry Collins playing QB for Penn State would mean fewer eight- or nine-man fronts to stop the run. While that should be easy to see on film, what might be less immediately obvious are defenders who can’t overcommit to the run from the snap, potentially giving Carter the extra split second needed to break free.8
Note: College Teammates and the Best NFL Running Backs Drafted Since 1984
I’ll get to the answer in the next paragraph, but first consider that there have been 56 RBs taken in the top 20 picks since 1984. Out of those 56 players, 48 played their final college season with at least one lineman or quarterback who was drafted that year or the following one in the top 224 picks. Only eight did not.
That brings us to the answer: Sanders, Smith, Faulk, and Peterson are four of those eight backs. Perhaps the four best backs picked in the last thirty years―if we put Peterson above LaDainian Tomlinson―each played their final college season without a single QB or lineman who was drafted that same year or the subsequent one.9 Other than one WR each, Sanders and Faulk essentially played their last college season with no elite offensive talent. The same was true of the back Sanders succeeded at Oklahoma St, who also made it to Canton.
So, just as great college QBs generally lead RBs and WR/TEs to be overvalued, these examples suggest that future superstars at RB sometimes can be found by picking players who succeeded in college without NFL talent surrounding them.
- I thank commenter Stuart for suggesting this approach in the comments to last week’s post. [↩]
- Carter averaged 3.3 yards on 227 carries over his first five injury-plagued seasons. [↩]
- Two Penn State halfbacks were drafted in 1996, as well. One of them was Stephen Michael Pitts, who went to Middletown High School South (NJ), a school that also graduated Knowshon Moreno and, only slightly less famously, me. [↩]
- As before, a player’s relevant teammates are taken to be those who get drafted in the same or following year. The regression predicts a player’s AV in the first five years by a higher-order polynomial in draft position. Using this regression to generate expected AV is similar to using Chase’s chart. To avoid adding complexity, I don’t adjust AV down to get Marginal AV. I then regress VAE on a set of year dummies, which account for changes in the value of passing and running over time, and the value of a player’s teammates at a position. [↩]
- p-values in all cases of no clear effect are above 0.30. For all reported effect sizes, the estimates have p-values less than 0.10. [↩]
- To be consistent across years, rounds 1-3 means the first 96 picks. [↩]
- No effect for QBs had a p-value below 0.30. [↩]
- If that’s right, we might expect that this bias towards RBs with excellent QBs would have gotten smaller over time as teams learn and have access to better data. And it looks like that might be accurate. If we split the sample, we get a bigger effect for 1984-1996 than for 1997-2009, although the difference is not big enough to know for sure that things have changed. [↩]
- If we go back one more year, we add Eric Dickerson to the list, who did have one lineman picked the year after him. [↩]