Saturday, September 18, 2010

Defense Wins Super Bowls: Not So Fast

Conventional wisdom says that defense wins Super Bowls, but history says that a good offense is just as important. An examination of the NFL Super Bowl champions shows that, with a couple of exceptions, Super Bowl winners had a better than average defense and offense, as measured by points allowed and scored(We could measure the productivity of a defense and offense on more than that, but it really all comes down to points in the end). Not a surprising result, but how do the two categories compare as indicators for success? The graphs below shows points per game scored and allowed for the Super Bowl winners versus the league averages over the years. Large gaps on both sides of the average indicate those teams that were dominant on both sides of the ball during the regular season, the two most dominant being the 1991 Redskins and the 1999 St. Louis Rams. Teams with small gaps on both sides of the average overachieved in the post season after a so-so regular season, the 2007 New York Giants being by far the best example. The ragged nature of the graph indicates a mixture of teams with dominant offenses, teams with dominant defenses and those that were balanced. Clearly, based on the regular season results, one cannot conclude that a stout defense was any more of a predictor for Super Bowl success than a productive offense.

Thursday, September 9, 2010

Will the New Orleans Saints repeat as Super Bowl champions?

Let's start with something simple and fun.  Ignoring the many variables that go into a detailed analysis of an NFL team's season forecast(strength of schedule, personnel movement, age, leadership, injuries, etc.), what does history say about the Saints' chance to repeat?

If we look at history of the NFL since 1966, there have been 8 repeat winners of the Super Bowl, with the Steelers doing it twice(1974-1975 and 1978-1979). No team has pulled off a three-peat. On only 3 occasions has the defending Super Bowl champion played and lost in the Super Bowl the following year. So if they get there again, history says they will win it. 13 times the Super Bowl winner has not made the playoffs the next year. This has happened to the Steelers three times. The remainder of the time(19), the Super Bowl champ has been eliminated in the league playoffs.

So, based solely on historical trends, it is highly probable(70%) that the Saints will make the playoffs in 2010 and they have a 19% chance of repeating as Super Bowl champions. What do you think?

Wednesday, September 8, 2010

Welcome to the Numberyard

By themselves, numbers don’t possess any meaning. Put them in context and they have the power to move us in all directions across the emotional spectrum. Take the number 100. If this is my score on the recent Chemistry exam, I would have a right to celebrate. The same cannot be said when I am told that this is the number of days that oil from the Deepwater well spilled unabated into the Gulf of Mexico. The truth is that numbers don’t lie. They are the stuff of measurement and, as such, they represent a kind of evidence.

In our American justice system, one cannot be found guilty of a crime without there being some sort of evidence supporting the charge. In the same way, the validity of a claim, theory or opinion is bolstered by supporting evidence, usually represented by one or more well-chosen measurements. This then begs the question: How do I determine a well-chosen measurement? For some, a well-chosen measurement is one that best supports their position, reminding us of the adage that economists use statistics like a drunk uses a lamppost: More for support than illumination. Better to choose a measurement that has a high level of acceptance amongst your audience, one that is agreed upon gets us closer to the “light of truth.” This will go along way in establishing that your evidence can be trusted.

The pursuit of evidence, thus, is a two step process: choosing the right things to measure and gathering the results.  So it's time to start stocking the Numberyard with some good quality evidence and have some fun in the process.  Your contributions are welcome.