In this week's Varsity Numbers, Bill Connelly takes a page out of baseball's playbook and attempts to isolate power from efficiency.
18 May 2004
by Aaron Schatz
One of my more controversial statements during this website's first season was my assertion that DeShaun Foster was one of the league's least effective running backs during 2003. The public perception was that Carolina's running game provided a one-two punch, but DVOA ranked Stephen Davis as a league-average back while Foster was down near the bottom of the league.
Foster's low DVOA recently came up in this discussion thread on the excellent fantasy football site footballguys.com, an attempt to list the most underrated and overrated running backs for fantasy football in 2004. DeShaun Foster, who is #36 in the site's expert rankings, is listed as the number one most underrated back:
LHUCKS: "Definitely not a RB2, but #36 would put him at the end of the RB3's... I don't think so. Mr. Foster is one of the most talented RBs in the league, and the ranking of #36 will be scoffed at within the first half of the NFL season... mark my words."
Now, there is the question of course of why this writer feels Stephen Davis is going to be replaced as the starter in Carolina -- even if he declines a bit with age, a RB-by-committee situation (bane of the fantasy player's existence) is more likely. But aside from that, two footballguys.com readers who are also Football Outsiders readers came forward to remind everyone of Foster's low DVOA:
abrecher: "DeShaun Foster is chronically overrated on this board. Check out the stats at Football Outsiders. They rate him the third-worst RB in the league last year (of those with 75+ carries)."
Tick: "I agree that those formulas aren't very applicable for FF for the most part, but I do think they tell a bit about how good an NFL player is... and they say that Foster wasn't a great NFL player last season, while Stephen Davis was good. So in this context (looking at whether Foster will supplant Davis this season), I think it applies."
The mention of our stats got this response:
Colin Dowling: "Are you sure you want to use a complex rankings system that has Rock Cartwright ranked 18th?"
Now, the rushing numbers for 2002 and 2003 are now updated using the new version of DVOA that I introduced last week, and Cartwright has dropped down to 26th, from 15.0% DVOA to 6.5% DVOA. But that's still unexpectedly high for the third-string back on a team not known for its running game.
Dowling's choice of Cartwright of an example of how strange the Football Outsiders rankings are has an unexpected bonus, however. I don't think Mr. Dowling realized this when he mentioned Cartwright, but it turns out Cartwright and Foster might be the two players in the NFL with the most superficially similar numbers -- yet dramatically different DVOA rankings:
So, it turns out that Rock Cartwright and DeShaun Foster give us a great opportunity to delve into the DVOA system and why it is a better measure of a player's true worth to his team (as opposed to your fantasy football team, which -- as footballguys.com commenter Jason Wood points out -- runs on completely different principles).
Now, there are caveats to this analysis. This is the regular season only, and Foster played better (and more memorably) in the playoffs. Yes, Foster was recovering from an injury, and his 2003 performance might not be reflective of his true ability. With those facts noted, though, let's compare Foster to Cartwright down-by-down to see why Cartwright was a much bigger contributor to Washington's offense than Foster was to Carolina's.
|First Down||DVOA||VOA||Carries||Yards||Yd/Car||Avg. To Go|
That last column may be unclear; it represents the average yards to go for another first down on each back's carries. In actual performance, there isn't much difference between these two on first down. But first down is where Cartwright's much harder schedule has its biggest impact. 29 of Cartwright's carries -- including 15 on first down -- came against our #2 ranked rush defense, Dallas. He had 42 total carries against top ten rush defenses. Foster had only 22 carries against rush defenses in our top ten -- 21 against #6 Detroit and one against #5 Jacksonville.
|Second Down||DVOA||VOA||Carries||Yards||Yd/Car||Avg. To Go|
Cartwright had 11 first downs and three touchdowns on second downs; Foster had only nine first downs, and no touchdowns. Cartwright had only two carries on second down that lost yardage; Foster had eight carries that lost yardage. On 11 second-and-short carries (1-3 yards to go), Cartwright had seven for first downs or touchdowns; on 10 second-and-short carries, Foster had only three first downs.
Some of these differences are due to usage patterns -- Foster had only two red zone carries on second down all year -- but remember, DVOA compares each play to the league average performance in that situation. Cartwright's second-and-goal touchdowns are compared to other second-and-goal situations, and are superior to league-average. Foster's many carries on second-and-long are compared to other carries on second-and-long, and are found wanting. To see exactly how it works, look about four paragraphs down for the "swell math appendix."
|Third Down||DVOA||VOA||Carries||Yards||Yd/Car||Avg. To Go||Fum. Lost|
Cartwright was better than Foster on third down, but not colossally better, at least when compared to league-average in the situations he faced. Cartwright had 13 first downs (and one touchdown) on third down, Foster had only eight, but that's understandable given that Cartwright carried the ball more often on very short yardage. I will note that Cartwright's DVOA takes a huge hit in the new version of the formula from his one lost fumble, which took place on 3rd-and-goal on the opposing two-yard line (ironically, Carolina was the opponent).
Oops, we almost forgot something:
|Fourth Down||DVOA||VOA||Carries||Yards||Yd/Car||Avg. To Go|
Foster didn't have a single carry on fourth down, but Cartwright was perfect in his five fourth down carries. I've left one of them off the above table, because it was a 22-yard gain on 4th-and-10 from the New Orleans 49-yard line, the last play of the half. New Orleans could care less if Cartwright ran for 48 yards as long as he didn't run for 49. (The play does count in Cartwright's total season DVOA -- I don't think it makes sense to go through and subjectively determine whether every single play should count or not.) Even if we don't consider this 4th-and-10 play, it is still impressive that Cartwright took the ball four times on 4th-and-1 and had four first downs.
Now we've considered every run by Rock Cartwright and DeShaun Foster in 2003, and shown how Cartwright was more successful than Foster on every down. I could run the numbers for different areas of the field and get the same results. So how did Foster end up with more yards than Cartwright? Simply put, he ran more often in situations where more yards did not mean more success. Here is a breakdown of Cartwright and Foster based on yards to go on each play. Notice how Cartwright had more success in each situation, but was used much more often in short-yardage situations, while Foster ran more often with seven or more yards to go:
|DeShaun Foster||Rock Cartwright|
|To Go||DVOA||Carries||Yards||Yd/Car||Avg. To Go||DVOA||Carries||Yards||Yd/Car||Avg. To Go|
*includes one lost fumble
Look, it is certainly possible that DeShaun Foster is better than Rock Cartwright. This is just one year of performance for each player, and a part-time year at that. But in 2003, Rock Cartwright was better than DeShaun Foster in pretty much every imaginable situation during the regular season.
That doesn't mean I want Cartwright on my fantasy team -- he's now backing up one of the league's top three backs, and, besides, all those short-yardage runs mean nothing in fantasy football unless they come on the goal line. In fact, according to this Washington Post article, there's a possibility that Cartwright won't even make the Redskins, since he's seen as the kind of "fullback-style" runner that isn't used in a Joe Gibbs offense. Gibbs is the genius, but I think keeping Ladell Betts over Cartwright would be a big mistake.
As for Foster, well, he could have a big season if last year was just an injury recovery year. And I suppose that it is possible that Davis breaks down and the Carolina running attack goes from being two-thirds Davis to being two-thirds Foster. But if last year represents Foster's true ability, Carolina is going to find that depending on Foster doesn't win them a lot of games. DVOA doesn't determine fantasy value, but it does show me that I'd rather have Stephen Davis on my team, and if John Fox agrees, you would rather have Stephen Davis on your fantasy team too.
To give an example of how the DVOA system works, here are all the runs for Cartwright and Foster on second down with 1-3 yards to go. The three numbers at the end of each line represent the value in "success points" used by the DVOA system to judge each play, the expected value of the league average in that situation (adjusted for opponent), and the difference. First downs are in italics, touchdowns in italics and bold.
|DeShaun Foster||Rock Cartwright|
With the new version of DVOA introduced last week, Cartwright gets extra credit for his touchdowns, but so do all the other running backs who take the ball on the one-yard line. That extra "success value" is represented in the number given for "expected value."