Player Points - Jonathan Stewart
By Chase Stuart
June 29th, 2010

It's possible that owners at the turn of their fantasy drafts will have the option of taking either DeAngelo Williams with their first pick, Jonathon Stewart with their third pick, or both. With Williams' average draft position hovering around 14th overall (RB 9) and Stewart's at around 33 (RB 16), it's at least reasonable to say that the owner with 12th pick in standard leagues has a chance to land both players. So, is there any reason they would want to do that?

There are several interesting, theoretical discussions you could have about drafting both Panthers running backs. Does it improve your team's odds of winning on a weekly basis? Does it increase your team's odds of winning your fantasy championship? Does it decrease your variance on a week to week basis? Does it decrease your variance on the season level? Is decreasing your variance a good thing? Are you now cut-off from drafting Steve Smith? Or is Steve Smith even more valuable to you than before?

I'm most interested in the question about variance. For starters, let's note that in 2009, Stewart played in 58 quarters and Williams played in 49 quarters, as noted in the Michael Turner spotlight. Williams missed three full games with an ankle injury, along with the last three quarters of the Vikings game due to the same. Stewart failed to receive a touch in each of six different games, but played in most of every game (or, at least, as much of "most" of a game as a typical second-string running back). Williams ranked as the 8th best running back in fantasy points per quarter in '09, while Stewart just slipped into the top twenty (and, of course, Stewart might not have slipped into the top 20 if Williams never got hurt). Carolina running backs, as a group, ranked as the 5th most productive set of runners in '09 (behind Baltimore, New Orleans, Miami and Minnesota) after finishing 2nd (behind New Orleans) in 2008. Williams and Stewart rushed for 2,351 yards in 2008 with 28 touchdowns, and then gained 2,250 yards and 17 rushing TDs last season. The biggest difference between '08 and '09 for the Panthers was the drop in play by Jake Delhomme; he ranked 4th in the league in yards per attempt in 2008 and threw only 12 interceptions in 16 games. But last year Delhomme imploded, ranking 25th in yards per attempt and nearly doubled his interception rate. While we don't know how well Matt Moore (or Jimmy Clausen) will play in 2010, the odds are that the passing attack will be at least slightly improved. And that might mean enough touchdowns for Stewart and Williams to be a viable combination this year.

So what should someone expect if they draft both players? Assuming both players are productive, will having them increase or decrease your team's variance? You can make arguments for both cases. For example, in week six against Tampa Bay last year -- the worst defense against the run in the NFL -- Williams rushed for 152 yards and 2 TDs while Stewart chipped in with 110 yards and a score on the ground. In that situation, it wasn't a coincidence that when Williams had a big game, so did Stewart. Williams had a great game because he was playing a terrible team with a terrible defense, a recipe for lots of production on the ground and a run-heavy game plan; those are two of the key ingredients in having your second-string running back put up big numbers, too. On the other hand, look what happened the Panthers traveled to New Orleans. Williams had a huge game(149 yards, 2 TDs) but Stewart was held to just 38 total yards and no scores. The Saints defense was pretty good, and obviously their offense was fantastic last season: that meant the Panthers couldn't run out the clock in the fourth quarter but rather were trying to catch up in a game they ultimately lost. Therefore, Stewart didn't get many touches or yards precisely because Williams was the main man that day.

But those are just two examples. What do the numbers usually say?

There have been forty-four pairs of teammates since the merger to play in at least 13 games and finish among the top 20 fantasy running backs in the same season.

Perhaps the most obvious example of the two player's fates being negatively correlated is that when one of them is injured, the value of the other skyrockets, as we saw at the end of last season. I calculated their weekly fantasy point scores for every week of their regular seasons, assigning scores of zero in weeks when a player was injured. I then checked the correlation coefficient between each running back's weekly fantasy scores against that of his teammate's; a positive correlation coefficient indicates a positive, direct relationship, meaning when one player does well, so does the other. A negative CC means the player's performances were negatively correlated; whether positive or negative, when the correlation coefficient is close to zero, that means the relationship isn't very strong; the closer the CC number is to 1 or -1, the stronger the relationship. The table below shows the results for all 44 pairs of running backs, with their year-end fantasy ranks in parentheses:

Team
Year
Running Back
Running Back
CC
CAR
2009
Jonathan Stewart (12)
DeAngelo Williams (15)
-0.41
MIN
2007
Adrian Peterson (5)
Chester Taylor (20)
-0.27
NOR
2006
Reggie Bush (10)
Deuce McAllister (15)
0.23
TAM
2001
Mike Alstott (19)
Warrick Dunn (20)
-0.18
TAM
1999
Mike Alstott (14)
Warrick Dunn (19)
0.46
TAM
1998
Warrick Dunn (17)
Mike Alstott (18)
0.30
MIA
1995
Bernie Parmalee (15)
Terry Kirby (17)
-0.24
MIA
1993
Terry Kirby (7)
Keith Byars (19)
0.04
SDG
1991
Ronnie Harmon (14)
Rod Bernstine (17)
0.13
CLE
1991
Kevin Mack (10)
Leroy Hoard (12)
-0.35
CHI
1990
Neal Anderson (3)
Brad Muster (17)
0.06
SEA
1990
Derrick Fenner (6)
John L. Williams (8)
-0.15
SEA
1988
John L. Williams (8)
Curt Warner (11)
0.35
CIN
1988
Ickey Woods (7)
James Brooks (9)
0.05
DEN
1986
Sammy Winder (17)
Gerald Willhite (20)
0.25
CLE
1985
Earnest Byner (12)
Kevin Mack (17)
-0.06
SFO
1984
Roger Craig (8)
Wendell Tyler (9)
-0.24
DAL
1983
Tony Dorsett (9)
Ron Springs (17)
-0.13
WAS
1983
John Riggins (4)
Joe Washington (20)
-0.35
DAL
1981
Tony Dorsett (11)
Ron Springs (20)
0.29
RAM
1980
Cullen Bryant (13)
Elvis Peacock (18)
0.10
GNB
1980
Eddie Lee Ivery (14)
Gerry Ellis (17)
-0.10
ATL
1980
William Andrews (7)
Lynn Cain (15)
-0.28
NOR
1979
Chuck Muncie (7)
Tony Galbreath (15)
-0.06
MIN
1978
Rickey Young (12)
Chuck Foreman (15)
0.58
CHI
1978
Walter Payton (1)
Roland Harper (10)
0.33
BAL
1977
Lydell Mitchell (2)
Don McCauley (20)
-0.05
CLE
1977
Greg Pruitt (6)
Cleo Miller (16)
-0.11
SFO
1977
Delvin Williams (10)
Wilbur Jackson (19)
-0.12
PIT
1976
Franco Harris (6)
Rocky Bleier (14)
-0.52
STL
1975
Terry Metcalf (5)
Jim Otis (13)
-0.48
BUF
1975
O.J. Simpson (1)
Jim Braxton (9)
-0.59
CIN
1973
Boobie Clark (5)
Essex Johnson (8)
0.26
OAK
1972
Charlie H. Smith (13)
Marv Hubbard (18)
0.12
DAL
1972
Calvin Hill (4)
Walt Garrison (9)
0.04
MIA
1972
Mercury Morris (8)
Larry Csonka (20)
-0.09
ATL
1972
Art Malone (3)
Dave Hampton (14)
-0.11
SFO
1971
Vic Washington (5)
Ken Willard (17)
0.05
MIA
1971
Larry Csonka (6)
Jim Kiick (20)
0.00
CLE
1971
Leroy Kelly (2)
Bo Scott (12)
-0.02
GNB
1971
Donny Anderson (10)
John Brockington (13)
-0.31
BAL
1971
Norm Bulaich (8)
Tom Matte (18)
-0.42
MIA
1970
Jim Kiick (7)
Larry Csonka (13)
0.45
CLE
1970
Bo Scott (4)
Leroy Kelly (11)
0.24
Average
-0.03

As you can see, the performances of Stewart and Williams in 2009 were negatively correlated; essentially, that game against Tampa Bay was an outlier. This is good if you're looking for a risk-averse starting lineup, because it means when one does poorly, the other usually does well. Perhaps the most obvious example of the two player's fates being negatively correlated is that when one of them is injured, the value of the other skyrockets, as we saw at the end of last season.

But note that, on average, there was almost no correlation between the scores of the two running backs. Nineteen of the 44 pairs had positive correlations, and only 8 of the 44 pairs had correlation coefficients of -0.30 or stronger. Obviously when an injury strikes, the value of the player's teammate increases, but it looks like generally speaking, there isn't much safety in numbers. If you like Williams or Stewart, draft one of them (or both). And if you want to draft Stewart to back up your investment of Williams, that's a fine idea. But I certainly wouldn't go out of my way to draft Stewart in the hopes of making my lineup more consistent.

Questions, suggestions and comments are always welcome to stuart@footballguys.com.