by Bill Barnwell
A few weeks ago, I was fortunate enough to attend the 2009 MIT Sloan Sports Analysis Conference in Cambridge. Not only did I get to put faces to the names of writers and analysts I've been reading for years (as well as meet several FO readers), but I was able to take in some very interesting presentations in the process.
One of those presentations was by Professor John Huizinga of the Chicago Booth School of Business, who worked with Sandy Weil on a paper called "Hot Hand or Hot Head?" The paper analyzes whether NBA players actually get a "hot hand"; essentially, whether a player is more likely to make a shot after making one previously, while also delving into which sort of players are more likely to think they have a hot hand.
It's a fascinating bit of research; for those who didn't attend the conference and are interested in Huizinga and Weil's findings, you can see their presentation here.
During their presentation, I started thinking about it in a football sense, and I was reminded of the essay that Shawn Krest did in the Ravens chapter of Pro Football Prospectus 2007, which looked at the idea of repeating carries on an individual level with yards per carry as the metric of analysis. The idea I see quoted more frequently is the idea that you can get a running back into a "rhythm" by giving him the ball multiple times in a row.
While Shawn's research was sound, there are some issues with the nature of running the ball multiple times and the situations teams were likely to find themselves in as a result that made me want to approach the dataset from a slightly different perspective.
What I did was take the entire league -- as opposed to the smaller subsets of individual players -- and simply track their DVOA on carries that occurred on plays where the previous play was not the player running the ball. That's listed as Carry 1, even if there were no subsequent carries. From then on, if the player carries the ball on the next play, it's Carry 2, Carry 3, Carry 4, and so on. For running backs, going anywhere past four consecutive carries plunges the sample below 70 carries to about 30 or so, so I'm only measuring up to four here.
In 2008, there was a clear and obvious trend. The DVOA of backs got worse with each carry, dropping off dramatically on Carries 3 and 4.
This was relatively similar to the trends we've seen looking back at previous seasons. From 2004 through 2007, there were only instances of backs improving, on average, their performance on a subsequent carry relative to the previous one, and it wasn't by very much. The overall trend is still that performance falls off a cliff.
Are there factors affecting this sort of research, too? Perhaps. Backs do tend to get repeated carries in short-yardage situations and close to the goal line, although there are plenty of instances of backs getting a first down on two carries and then running the ball from first-and-ten on a third carry. Backs getting consecutive carries are also more likely to be winning, which can indicate that their offense is better than league-average, or that the situation has deteriorated to the point where they're running the ball into the line and neither team cares to a level that DVOA can comprehend.
That mitigates the findings of this sort of research to an extent, but doesn't disqualify it. Running the ball with one player repeatedly to get him and your offense into a "rhythm" is only likely to put them on the sidelines.
On a final note, the most entertaining thing I discovered doing this research was from a Raiders-Jaguars game in Week 17 of the 2004 season. Check out the fourth quarter play-by-play. If I was Zack Crockett, I wouldn't have talked to Roland Williams for years. Or Norv Turner, for that matter. Ouch.