Division Odds, Week 8

The Saints are now favored to win the division, at least according to the Football Outsiders playoff odds simulation. The latest simulation has the Saints odds of winning the division at 63.8%, the highest they’ve been all year and the highest of any team in the NFL that doesn’t currently or formerly employ Peyton Manning. Check out the graph:

divisionOdds

Now, you might be wondering how the playoff odds fluctuate so much from week-to-week. How can the Falcons go from 60% likely to win the division in week 4 down to 10% likely in week 8. Do things change that much?

The brief answer is that while the quality of the teams don’t change that much, our understanding of how good or bad they are does. The NFL has an impossibly small sample size of 17 weeks, which means it’s really hard to tell how good NFL teams actually are. You wouldn’t try to judge a baseball batter after 17 at bats, right? But that’s what we’re trying to do here*. So after each week, we understand a little more about the teams, and our judgment of how good or bad they are changes to reflect this new understanding.

*I get that many of the advanced stats use individual play data to expand their sample size, but my general point is valid, especially looking at overall team quality.

So when Football Outsiders does their playoff odds simulations, they use their judgments of how good or bad teams are (primarily their DVOA ratings) in the simulations. Each week, we get more information, so the odds can shift dramatically.

An imperfect metaphor: think of the playoff odds simulations (and other advanced stats, really) as a complex version of Battleship in which your opponent’s ships represents the actual quality of an NFL team. The extra challenge: you don’t know where opponent’s ships’ size, location, or shape. And your opponent is probably a cheater.

Each week, Football Outsiders calculates the playoff odds based on the information they have so far (DVOA and the team records). Each simulation is like a shot in Battleship: a guess as to the quality of the team (in this metaphor: the location of your opponent’s ship). The following week’s game suggests whether or not the last shot was a hit or a miss: did the team do as well as expected (hit!), or did they do more poorly (miss)? Then they re-calculate the team quality (DVOA) and re-calculate the playoff odds, taking the prior week’s performances into account, and “shoot” again. This time with slightly more accurate information.

On top of all this, the actual wins and losses from each week make it easier or harder for teams to make the playoffs, changing the playing field entirely.

Failed metaphors aside, as hits and misses accumulate, you get a better idea of the ships’ locations and the teams’ quality, which makes your later shots more likely to hit than your earlier shots. Given the tremendously small sample (7–8 games so far), it’s not surprising that the odds shift so dramatically, especially early in the season. That’s why I like to look for general trends and changes rather than focusing on a specific week’s data point. I think it’s more accurate to do so.

That said: 63.8%? [That’s worth singing about][6].

[6]: http://www.youtube.com/watch?v=VXkVZ0rloio