I am continually amazed by how people think. Every year, "experts" do their best to preview the upcoming season, and every year they are completely surprised by how certain teams perform. Last year the over/under for the number of Chiefs wins was 6 despite the easy schedule, and I saw many people pick the under. Seemingly everyone thought that the Redskins would be at least average and that the Cowboys would compete for the Super Bowl. Of course, the experts turned out to be completely wrong and were shocked at how well the Chiefs and Bucs did and how poorly the Cowboys and Vikings did. So they took a close look at how they evaluate teams and tried to correct their errors, right? Of course they didn't, but not to fear, I actually have done a little research into this, and unlike the "experts" actually gave it some thought, so hopefully we will be at least somewhat accurate in predicting the next season.
One of the most important things the "experts" don't understand is the sheer amount of change from one season to the next. Take for example, this graph:
If the "experts" saw this their heads would explode. The R-squared measures how much the two variables are correlated with each other. It's a number between 0 and 1. .05 is very poor, much lower than you would expect. It basically means that how much you win one year has very little to do with how much you win the next year.
But this actually isn't true, because of another mistake almost everybody makes when doing previews (and I even made this mistake when I first did that graph). The important thing is the change, which gets surprisingly little respect. If the Carolina Panthers traded for Brady or Manning, they would improve much more than if the Chargers did the same, because it would make a much bigger change for the Panthers than the Chargers. But this rather obvious fact is surprisingly overlooked. The Chiefs radically improve their WR2 spot with Baldwin and Breaston and nobody pays attention, but when the Falcons replace a not so bad WR2 with a rookie and everybody loses their shit. Even if Baldwin and Breaston aren't as good as Jones, our improvement was greater because our WR2 spot was so much worse than the Falcons last year.
So when we redo that same graph from above, but instead focus on how wins one season effect the change in wins, we suddenly get a very meaningful result.
The R-squared is now .43, which is very, very good for this type of analysis. But the fact that the line of best fit crosses the x-axis at exactly 8 wins speaks volumes about the quality of this correlation (you can check that intersection point yourself using the equation above the R-squared).
The reason behind this effect is obvious. The higher you are, the further you have to fall and vise versa. Yet the "experts" will be shocked when the Falcons and Patriots don't tear up the league and the Panthers win more than 2 or 3 games.
Another reason for that effect is that there is a lot of luck in football, or at least a lot more than the "experts" think. I'm amazed that people aren't talking about how lucky the Packers were to win the Super Bowl, considering that had the Bucs beat the Lions they wouldn't have even made the playoffs. Yet this 10-win team with a history of being injury prone is thought to be the best team in the NFL by all the "experts".
(I'm skeptical of NFC Champion teams in general. Consider the fact that 63% of the NFC has been to the Super Bowl in the last ten years, while only four teams in the AFC have gone in that time, and it would be three if it wasn't for the Raiders in 2002. The NFC teams have a built in advantage over AFC teams because the AFC teams have had 3 dynasties to deal with in the last couple years (Colts, Patriots, and Steelers) while the NFC has had a bunch of nobodies. But since the better team doesn't always win, we get an occasional NFC Super Bowl champion who was actually inferior to the team they beat.)
Consider turnovers, which are probably one of the biggest areas of luck in football. Sure, a team can do a little to effect turnovers, but ultimately you don't have a whole lot of control over them. This graph plots change in wins from 2009 to 2010 against turnover differential in 2009 (Note: I have a couple errors in here as point out by Jaywalk3r. Fortunately fixing my errors makes the correlation stronger, so I'm not too motivated to go through the trouble to fix them.)
(Last time I posted this there were some questions about how "good" this correlation was, since an R-squared of .1 is good but not great and I only had 32 data points. Jaywalk3r ran some numbers for me and found out that there's only a 3% chance this correlation occurred by chance, which is good enough for most to consider it significant.)
I think this graph illustrates one of the better predictors that is commonly ignored. Simply put, with something like turnovers, teams will regress to the mean, so that a team like the Falcons, who's 13-win season was built on a high turnover margin, have a very good chance of being disappointments. (The Chiefs also had a high turnover differential, but considering that we have historically been the best team in the NFL in turnover differential by a wide margin, I'm not too concerned.)
With this in mind, the ESPN Power Rankings look hopelessly naive. Every single team with a winning record from last year is ranked above every single team with a losing record. It gets even better when you look at the details, where they say the three toughest divisions are all in the NFC, despite the fact that the NFC hasn't had a winning record against the AFC since the 90s (another fact that I'm surprised nobody talks about). The AFC East went 9-7 against the NFC North, the AFC North also went 9-7 against the NFC South, and the AFC South, the weakest division in the AFC by wins, managed 7-9 against the NFC East. And yet these three divisions are supposed to be the toughest in football.
NFL "experts" get paid to know more about football than everybody else, yet they almost always fail to do their jobs. Then again, maybe it's not so surprising that those who are paid to be knowledgeable actually have no idea what they're doing. It wouldn't be the first time a highly paid person failed to do their job.