Writers of Pro Football Prospectus 2008

23 Jan 2009

Smarter Stats: The Worst Winners

This week's Washington Post column uses Pro Football Reference's Expected Wins, which does a Pythagorean projection for all Super Bowl teams going back to the Packers and Chiefs in 1966, to spot the five biggest disparities between Super Bowl winners with lower Estimated Wins than the teams they beat. The biggest gap happened last year (gosh, what a surprise), and the Colts-Bears Super Bowl is also included in the top five. Based on Arizona's and Pittsburgh's regular seasons, a Cardinals win in XLIII would be the second-biggest Pythag upset in Super Bowl history. Yeah, there's a recent trend going on here.

Posted by: Doug Farrar on 23 Jan 2009

7 comments, Last at 27 Jan 2009, 8:42pm by chappy


by Tim (not verified) :: Fri, 01/23/2009 - 10:20am

Is there reason to believe there is something fundamental that has changed regarding the league and/or the playoffs that would lead to the recent trend (e.g., 4 division alignment), or is it possible that the calculations are becoming less valuable as predictors?

by joepinion (not verified) :: Fri, 01/23/2009 - 11:25am

"The stat is based on the Pythagorean Theorem, which was invented by Bill James about a generation ago"

I know that's not supposed to be a joke but it's still funny.

It bugs me that you guys don't insert statements that at least leave open the possibility that something as simple as points scored over a 16-gave season isn't a meaningful enough stat. You just throw your hands up in the air and say "ahhh, whoever gets hot!"

For one thing, the Cardinals haven't won the Super Bowl yet. Maybe they'll lose by 30.

For another thing, just as the playoffs are only 11 games, the Arizona Cardinals regular season is ONLY 16 GAMES. Also, the solution is staring us right in the face: The Cardinals dogged it for about a month and got crushed.

by chappy (not verified) :: Fri, 01/23/2009 - 4:44pm

What about some adjustment for strength of schedule here. Sure it's much less of an issue in baseball, but certainly not in football. For example, it is ridiculous to put the Jets/Colts so low. The Jets literally played in a different league back then (with 6 fewer teams in the AFL!?) with a different mix of team quality.

My hunch is that this year's SuperBowl is an even bigger mismatch than purported in this article, simply due to the fact that the NFC West was so weak this year. Nearly 6 of their 8 wins were guaranteed by their division. I would bet the opposite is true for Pittsburgh--especially since they played the NFC East for their out of conference games.

by Jacob Stevens (not verified) :: Fri, 01/23/2009 - 7:15pm

"...especially since they played the NFC East for their out of conference games."

So did the Cardinals. And the AFC East (three teams with winning records) and the Panthers and Vikings. They played more teams with winning records than losing records.

DVOA does rank the Steelers as having a much stronger schedule for the year, but I'm not comparing, I'm just contending Arizona's strength of schedule wasn't quite as soft as the common consensus would have you believe.

Hell, even DVOA is fallacious in that regard. The AFC East was significantly docked for facing the NFC West, who was definitely weak, but whose own strength of schedule was docked, aside from just being a weak division, for facing the AFC East, whose DVOA was more significantly lower than nearly any division's VOA, in large part on account of facing the NFC West.

It's like, the worse a division gets in true aptitude on the field, the logarithmically more their DVOA decreases, like some sort of black hole of suck, and vice versa for a couple of strong teams facing a few other strong teams. The DVOA theory of general relativity? Is beating Pittsburgh outside of the Arizona Cardinals' light cone? OK, this post is going no where fast, so I'll stop now, but just remember that the Steelers aren't facing the regular season Cardinals, on Sunday.

by chappy (not verified) :: Tue, 01/27/2009 - 8:42pm

Um, not sure what you're saying here. The Cardinals had the 27th hardest schedule and Pittsburgh had the 4th hardest. According to the numbers on the site, this is a 10% difference in strength of schedule. The theoretical maximum SOS difference match-up would have been Cleveland (1st) versus Buffalo (32nd) at about a 16% difference. I don't discount that the Cardinals might be better than their regular season numbers or that the Cardinals can't win the SB, but this is really a mismatch along a number of metrics.

by Jacob Stevens (not verified) :: Fri, 01/23/2009 - 7:28pm

Doug, it surely isn't on the same scale as those improbably Super Bowl runs, but since the past two winners are on there, the one prior to them also seems widely attributed more to the team just getting white hot in the postseason (if not quite in the Super Bowl) than the team being a natural favorite to run the table (who I reckon woulda been the Colts).

Naturally there's a overwhelming selection bias in "hot" teams in the playoffs winning the championship, but the frequent example of a team playing above their regular season level is remarkable.

Still feels more like a remarkable coincidence, though, like 0-16 following 16-0, than anything tangible. After all, none of the teams have yet been able to reproduce it.

by tally :: Mon, 01/26/2009 - 5:37pm

I believe the correlation coefficient of Pythagorean projection to wins and wins in the following year is at least equal to DVOA, and better than anything else (such as total wins), so there is some predictive value to it.