The "Patriots effect"
One of the things that has been anecdotally observed over the past several years is what has sometimes been called the "Patriots effect." That is, the Patriots are a tough, physical team, well-coached, disciplined and smart, and playing against them seems to take something extra out of a team. They play worse the weekend following a game against New England than they do in general, either because they have been physically worn down or because the Patriots have showed the opposition how to deal with some aspect of their game.
But it was, as near as I can tell, just anecdotal, observational. I decided to spend a few minutes trying to quantify it for the 2007 season.
The Patriots have played 14 times against 11* teams that played a game the following week**. This table shows how Patriots opposition did overall, against New England, against everyone except New England, and in the week following the Patriots game.
|Against New England||0||15||239||551||15.93||36.73||0|
|Against Everyone Else||109||104||4611||4282||21.65||20.1||0.512|
|Week after facing New England||4||10||284||351||20.29||25.07||0.286|
Offensively, there isn't much difference. In terms of winning percentage and defensively, there are big differences. Against non-New England competition, these teams allowed 20 points per game and compiled a .512 winning percentage - the week after facing the Patriots, they allowed 25 points per game and compiled a .286 winning percentage. And of the four wins, three were against two of the worst teams in the NFL (the Bills and Redskins beat the Jets, the Browns beat the Dolphins.)
Obviously, we're talking about very small sample sizes here. But we can look at a couple of different statistical tests for significance. First, the winning percentage. If that set of teams has a .512 winning percentage, what are the odds of winning just 4 games out of 14? Simple binomial probability suggests that teams with a .512 winning percentage have about a 7.6% chance of winning fewer than 5 out of 14. So it is possible that only chance and schedule effects are in play here. At a 5% significance level, we can't say that it isn't just luck, at a 10% significance level, we can.
Looking at the defensive numbers, we can do a t-test, and it suggests that, at the 5% significance level, the difference is statistically significant.
Bottom line - we cannot say for sure that there is a "Patriots effect," but there is evidence that supports it fairly strongly, at least for 2007. And it will play into my playoff predictions this weekend (I'm looking at you, NY Giants, and you, Jacksonville Jaguars...)
* - They played the Jets, Dolphins and Bills twice each.
** - Cincinnati had a bye following the Patriots game, the Giants have a playoff game this coming weekend.