Saturday, August 28, 2010

NFL seasons have become less predictable over time

With the NFL regular season a couple of weeks from commencement, predictions for 2010 are about to (dis)prove their worth. The prevailing mindset seems to be that a few surprises are inevitably in store, as has more-or-less been the case since the Colts clinched a first-round bye in the playoffs during the 1999 season, a year after finishing 3-13. Last year, of course, the Saints won it all after a decade of mediocrity.

How relevant does the idea of parity in the NFL remain relative to decades past? When the term "dynasty" is used, how far back in time must one go to apply it meaningfully? The Patriots of the early 2000s? The Cowboys of the early and mid-nineties? The 49ers of the eighties? Or all the way back to the Steelers of '74-'79, who won four Superbowls in six years? It's a question sports buffs could argue over indefinitely.

As I don't have a compelling qualitative case to make, I'll defer to the numbers and take the quantitative approach. The following graph shows the season-to-season wins correlation over time. That is, the r-value for the 2008-2009 seasons is computed by taking each team's number of regular season wins in 2008 and correlating them with the number of regular season wins each time had in 2009. The individual data points are provided in a table at the end of this post.

There are some expected year-to-year random fluctuations, but three broad periods emerge, each one successively defined less by steady performance season after season than the period before it. In the seventies, teams' regular season records consistently correlated at higher than .50. If you were a Steelers fan in 1974, you had strong reason to believe that the black and gold would have a very real shot at hoisting up the Lombardi trophy again next year. That's in contrast to the state of the game today--who is seriously predicting a repeat performance by the Saints in Arlington next February? (The odds-makers show them tied with the Cowboys as the second most likely team to win, behind the Colts; I'm highly skeptical).

After a couple of topsy-turvy years ahead of the players' association strike that severely disrupted the 1982 season, the 49ers helped bring a brief return to seventies-era consistency. By the late-eighties, the capriciousness of the Bengals (who had only won 4 games a year before Boomer Esiason came up a little short against Joe Montana in Superbowl XXIII) became relatively commonplace.

Year-to-year performance would hover around .40 from that period through the end of the nineties. Then, in 1999, pandemonium ensued! After leading the 49ers to a 3-1 start, Steve Young's career would come to an end at the hands, er, shoulders of Aeneas Williams, one of the greatest defensive backs in NFL history (I realize that assertion is controversial, but Young never played again, so it effectively did so). San Francisco would go on to lose all but one of their remaining 12 games, ending 4-12 a year after they had gone 12-4. The Colts, meanwhile, did the opposite, as Peyton Manning, in the manner of Troy Aikman a decade before (and then some), exploded onto the scene in his second year as a pro, leading the Colts to a 13-3 regular season after his disappointing 3-13 rookie year.

This is about the time "parity" became cliche (although it more accurately described perceived randomness rather than equality in teams' performance on 'any given Sunday'). No season before or after has ever outdone 1999 in jumbling things up from the previous year, but excepting 2008-2009, which was unusually consistent, the year-to-year correlation has floated around .25 ever since, repeatedly failing to constitute a statistically significant result at even 90% confidence.



silly girl said...

Off topic.

AE have you ever done a post on the total number of SAT takers with a score of X or high per year?

I want to know if the total number with say Math 600 or higher is increasing, decreasing or stable. I couldn't find anything and wondered if you had seen or done such a search.


OneSTDV said...

The odds-makers show them tied with the Cowboys as the second most likely team to win, behind the Colts; I'm highly skeptical

Oddsmakers don't actually make the odds reflective of who will win - they make the odds so that the most people will bet. That's why easy choices like the defending champs, Colts, and America's Team (Cowboys) are always up high. Not everyone betting is that informed so by giving these teams the highest odds, that suckers people into betting more.

Of course, even that doesn't really make sense because people should bet based on cost-benefit analysis and high odds from the oddsmakers doesn't really equate to this. But gambling has never been a logical pursuit.

dk said...

"That's why easy choices like the defending champs, Colts, and America's Team (Cowboys) are always up high. Not everyone betting is that informed so by giving these teams the highest odds, that suckers people into betting more."

Then you should go to matchbook and sell/short contracts(to win SB) on the saints and colts. I'd think twice though. There are extremely large and liquid sportsbetting conglomerates that move the lines on these games in vegas. And I assure you they crunch just as many numbers as you do. It's a very efficent market.

I have fun with MB, lowest commisions out there, cheap fun. But I have no illusions that I'm going to beats the oddsmakers.

Audacious Epigone said...

Silly girl,

I haven't really poked around for a site with absolute numbers for SAT scores over time, but I bet someone in the Steveosphere has.

silly girl said...

My curiosity got the better of me. I totaled it up myself from the college board archives. I emailed it to Sailer with my questions and comments.

Math SAT 600+

1996 - 246,398
1997 - 264,149 +7.2%
1998 - 277,674 +5.1%
1999 - 289,198 +4.1%
2001 - 312,533 +8.1%
2002 - 335,339 +7.3%
2003 - 367,924 +9.1%
2004 - 362,000 -1.6%
2005 - 391,436 +8.1%
2006 - 378,632 -3.3%
2007 - 366,234 -3.3%
2008 - 379,493 +3.6%
2009 - 392,989 +3.6%

Average yearly increase 1996-2004 = 5.6%
Average yearly increase 2002-2009 = 2.3%