Saturday, 2 February 2013

On The Defensive

The climax of the NFL season is upon us, with Super Bowl XLVII (47) taking place in New Orleans tomorrow afternoon / evening. The San Francisco 49ers represent the NFC and the Baltimore Ravens the AFC, and Nate Silver of US Election fame has taken a look at this game from a statistical standpoint for the New York Times. Thanks to Scott for forwarding me the link.
Does defense win championships? Stat-geek sports fans like me tend to distrust this old saying. Scoring a point helps just as much as allowing one hurts. And in football, the proposition risks ignoring the role played by the likes of Tom Brady. It is the case, however, that the better defensive team has usually won theSuper Bowl — and done so far more consistently than offensive juggernauts. The Web site Pro Football Reference publishes a statistic called the Simple Rating System (S.R.S.), which evaluates each team’s offense and defense based on the number of points it scored and allowed relative to the league average and adjusted for strength of schedule. Of the 92 teams to have played in the Super Bowl before this year, I identified those with the 20 best defensive and offensive ratings, according to S.R.S. (see charts above).
The defensive list contains teams that you would expect, like the 1985 Bears. These teams have compiled a 14-6 record (.700) in the Super Bowl. Their winning percentage is actually nearly 80 percent when you ignore the three cases, Super Bowls IV, VIII and XLV, when two of the all-time great defensive teams faced each other.
The 20 best offensive teams, however, are just 10-10 in the Super Bowl. There have been successes in this group, like the Saints under Drew Brees, but there have been just as many failures, including two of Brady’s Patriots teams. (During his three championships, the Patriots had a much better balance between offense and defense.)
The reasons that exceptional defenses fare so much better in the Super Bowl are still somewhat murky, but this factor bodes well for this year’s 49ers, whose defense belongs in the elite group, according to S.R.S. (it ranks 17th among Super Bowl teams). The Ravens, despite all the hype surrounding Ray Lewis, allowed a rather pedestrian 21.5 points per game this year.
The 49ers also have the better offense, according to S.R.S., so there isn’t much to recommend the Ravens . . . unless you look at the more sophisticated rankings published by Football Outsiders. Their system, known as Defense-adjusted Value Over Average (D.V.O.A.), accounts for a team’s success or failure on every play it ran during the year and not just on final scores.
Those rankings find that while the 49ers had the better offense and defense, the Ravens had the best special teams in the league this year. If they do pull off the upset, on the heels of Steve Weatherford’s game-changing performance for the Giants in last year’s Super Bowl, perhaps it will be time for a new clichĂ©: punters win championships. But don’t count on that.


AL said...

Im not sure we can take seriously anything Nate writes now that his reputation is in the mud when he was proved wrong recently... saying the Niners won't be in the Superbowl.

On another note, is it possible to beat variance?

WhyAlwaysMe said...

AL if you bought Nate's book you would learn there is no right or wrong, only probabilistic forecasts. Bayes' theorem innit.

Why would you want to beat variance??? Variance is variance. It just is.

Rich said...


i've resumed blogging after a couple years break. would appreciate a link as don;t think anyone follows me any more after all this time! thanks

Anonymous said...


have resumed blogging after a couple years off. would appreciate a listing as i don;t think anyone is following me any more after all this time! thanks

AL Little said...

I will check out his book WhyAlwaysMe.

Do you not think that predicting the winner of the election was far easier than say the winner of the superbowl which he couldn't do?

If variance just is, then surely it is not possible to find value over a long enough time period?

WhyAlwaysMe said...

Yes do check out the book it's well worth it.

Nate Silver did not predict the winner of the US election. What he did was build a model that predicted the most likely winner. Really his model was not genius, he was infact exposing the rather weak models of others, which was itself a big surprise.

Regarding the superbowl his model would have again be a prediction of the most likely winner. If he went 70% on the 49er's that still leaves Baltimore with a 30% chance. He is only coming up with two sets of numbers and not saying who will win as such. It is not about being a hero or villian but an attempt to build an accurate model.

Value and variance are not related to each other. You get value by beating the closing line or by having greater knowledge than the market. Variance can be reduced by altering your bets. Asian handicaps or point spreads are a good example of this. If you are betting at odds below 2.0 your variance will be much reduced and it will be quicker to tell if your system is working.

Variance is nothing to worry about, you just have to flex your bank to it that's all. Identifying value is the real problem!

AL said...

cheers WhyAlwaysMe.

some of my bets are at 100/1+ so variance can be an issue for me. kitty is bulging but i am worried about variance.