18 Oct 2007
In the Extra Point thread discussing the merits (ahem) of Roger Goodell's master plan to put radios in all players' helmets, Brian asked:
Brian: Do we even have any statistical proof that offensive efficiency drops league-wide when teams are on the road?
Stuart Fraser: We have half the answer already -- in the St. Louis chapter of PFP 2007, Bill and Aaron investigated false start penalties, showing that teams accrued more of them on the road and in domes, and Bill Moore showed the previous year that some stadiums do draw more false starts than others, with the Edward Jones Dome in St. Louis, Lincoln Financial Field in Philadelphia and Buffalo's Ralph Wilson Stadium leading the way.
But what about the plays that aren't blown dead? Are those affected by backs or linemen not hearing assignments clearly, teams having to go to a silent snap count, diminished ability to audible or other problems created by crowd noise?
We can get some idea of how teams are affected by comparing home and road DVOA. There are two possible approaches -- comparing home and road offensive DVOA to see which teams are most affected by the various disadvantages being on the road brings, or comparing home and road defensive DVOA to see which stadia disrupt opposing offenses the most. I chose the latter, looking for the greatest differences between home and road defensive DVOA (these stats are available in the premium database) for all seasons 1996-2006. This gives a total of 342 team-seasons. I didn't look at this year as the sample size is smaller and the opponent adjustments aren't yet at full strength, so it wouldn't be entirely comparing like with like.
Now, crowd noise isn't the only factor in home advantage, and it's difficult to separate any one factor from any of the others, which is something to keep in mind whilst reading the rest of this article. The statistic I'll be using is "home DVOA" - "road DVOA," and as this is defensive DVOA a better number is negative. Therefore, the stronger the home advantage, the more impaired an opposing offense, the more negative the difference between home and road should be.
The mean difference between home and road defensive DVOA, counting every team for every year, is -7.2% -- that is, offenses are an average of just over 7 percent less effective on the road than they are at home. Dome teams (including Dallas, as Texas Stadium's architecture is built in the style of a dome) fare slightly better than league average, with defensive efficiency up by 8.4%.
There is a lot of variation between teams and years -- the standard deviation is 14.6%. Given this, it's hard to say anything with confidence about the individual teams or stadia (for which the sample size is at most 11), but here's the list of top ten mean home advantages, just for fun:
Some of these are names we expected -- the Rams and Bills show up as they did for false starts, Arrowhead is famously noisy, and Minnesota, Indianapolis and Detroit also play in noise-enhancing domes. But Arizona? The Cardinals actually have the greatest single-season difference between home and road defensive DVOA, in 2003, when they posted a respectable -4.6% at Sun Devil Stadium and an, um, less respectable 40.8% on the road. It's not just that year, though -- take it away and they'd still edge out the Lions for seventh. I guess all those empty seats must have been really intimidating. The safer conclusion might be that home advantage isn't just about crowd noise, as the weather in Arizona is probably a factor, though Denver is middle-of-the-pack with a more or less league-average -6.2% advantage.
To close our visit to small sample size theatre, I'd like to honor the players and fans of the Houston Texans, who have combined to produce a team that actually manages to defend better (in four of the franchise's five completed seasons) whilst playing on the road.
So, to summarize -- yes, there is a quantifiable drop in offensive efficiency for road teams (or rise in defensive efficiency for the home teams); it's slightly larger in a dome, suggesting crowd noise is at least a factor; and small sample sizes produce wacky results.
12 comments, Last at 20 Oct 2007, 10:32pm by bigmaq