The Suuuuuper Bowl

Tomorrow is Super Bowl Sunday, a once a year American special. The matchup between Green Bay and Pittsburgh manages to combine two of the best three teams in football (New England just didn't show up against the Jets). In fact, the Packers and the Steelers have been at the top since the very beginning of the season. After week 3 I ran the rankings for the first time. Number 1 was Pittsburgh, number 3 was Green Bay. Both teams were among the four that spent time on top throughout the season.

Weeks 3-6: Pittsburgh

Weeks 7-9: Tennessee

Weeks 10-13: Green Bay

Weeks 14-17: New England

Okay, I don't understand Tennessee either (some mid season turmoil really threw them off), but the other three numbers ones showcased the power of predictive systems.

Points line for the game gives the following prediction:

Green Bay 22 Pittsburgh 21

The expected total points is 42.76, and the expected differential is 1.43. By the rankings system the Packers have a 52.4 percent chance of winning, but with a little twist we can also calculate the chance of victory. Calculating the game score standard deviation (and adding in uncertainty on each team's value) then averaging the two we get the expected standard error for the super bowl is 11.50 points. Importantly, it's much larger than the expected differential. A z score of .124 means the Packers should win 54.9% of the time. Interestingly, with a line of 3 for the Packers, that means there's a 55.5% chance that the Steelers cover (umm, still not endorsing gambling no matter how good the Prediction Tracker thinks I'd be). Actually though, even with the 1.43 number there's a high degree of uncertainty because each team's rating has an error of about 2.5. On the whole what it all means is that we have two of the absolute best teams in football, and picking the winner might as well be a coin flip. It's gonna be a good game.