There’s an interesting article over at Harvard Sports Analysis looking at if the NFL is experiencing an unusual amount of parity this year. I thought we’d take some looks at that this week to help spice up the power rankings.
First, I think they are perfectly correct to sidestep using records as the best measure of team quality. Weird things happen to switch game results, there are only so many options after 6 or 7 games for each team, and broadly speaking you’d rather use something like point differential anyway. So kudos to them on their first choice.
However, making a couple of graphs and then eyeballing them does not make for quality analysis. They present the distribution of point differentials for this season and then the one for the prior ten years combined. Over ten years, you see a pretty nice normal distribution centered at just above 0. That is, more teams tend to be just above 0 than just below, but the other bins are pretty symmetric. Looking at this year, you get something normal-ish but noisy. The author concludes we are indeed seeing more parity because there are no very good teams, like the 2007 Patriots. I think that’s wrong.
First reason why: he says there are the same number of poor teams, which is incorrect. The worst teams so far this year, four of them, are in the -50ish to -100ish bin. Looking at the ten-year sample, there are fewer than ten historical teams who have been worse than that at the same point (I’m eyeballing the number from the graph). But there are over 300 teams in that sample. Similarly, there are fewer than ten historical teams better than the +50ish to +100ish bin, and no teams this year. So what we see is that historically speaking, 3% or less of teams are better than 100ish and another 3% or less are worse than -100ish. With 32 teams playing in a single year, that means we would expect to see maybe 1 team on either side. Not seeing that team this year doesn’t mean we have more parity.
A more formal way to do this is with a chi-square test. The chi-square basically asks if the frequencies in two distributions came from the same population; in this case, are the frequencies of the point distributions for this year and the past ten years the same. I eyeballed the numbers for each bin from the graphs and ran the test and the result is not near significant. That is, this year’s results are not different enough to say that we have parity or some other shift in distribution compared to previous seasons.
Another way to look for parity would be to use something more sensitive than point differential, like my power rankings. They use a variety of numbers from each team to evaluate how well that team is playing. Sometimes teams get points in unusual or unlikely-to-repeat ways, and these rankings partially adjust for that. Here’s how they look after last weekend.
We have a switch at the top compared to last week. The Bears were unimpressive enough in their win over the Lions to drop from number 1 to number 5, although they still aren’t that far off from the teams above them. Houston, fresh off delivering a beating to the Ravens, have reclaimed the number 1 spot.
From these ratings we see a spread from .17 to 1.75, more or less. I’ve been doing the power rankings for two years, so we can compare this year to 2011 and 2010. At this point last year the best team in the league was Green Bay followed ridiculously closely by the Patriots, both at a rating of over 2. The worst team was the Rams at -.11. In 2010 we had Pittsburgh up at around 2.2 and Carolina down at .25. So it appears that the low end of the league is still in a reasonable range, but we are missing those high-end teams. However, the average performance has also dropped this year. It’s about three tenths of a point below 2011 and 2010. If we were to bump everyone up those .3 points, the Texans would be just over 2 and not that far off from the best teams of the last couple years. So I don’t think we’ve seen any huge increase in parity this year compared to the past.
Alright, so how about moving forward? Let’s look at the season projections.
Obviously the best teams look to do well in the future. Houston, Chicago, and Atlanta all look to get to around 12 wins. No one projects to fewer than 4 wins either, which isn’t too shabby. Following on last week’s list of teams who need to turn things around (those projected to 7 wins or less), this week we have Tennessee, Jacksonville, Kansas City, Carolina, Oakland, Cleveland, the Rams (right at 7), New Orleans, and, sadly, Detroit. Dallas and Indy have dragged themselves just over 7, so I’m putting them in the ‘still have a shot’ category. These other teams have a lot of work to do.
Sport Skeptic 