Hawk Blogger 2009 Power Rankings Week 1

Power rankings are always debatable. I don’t buy into the gut feel methods most places use to determine their rankings, so I developed a formula a few years back that attempts to take at least some of the subjectivity out of the discussion. My approach is simple, I measure offensive and defensive efficiency based on the Yards Per Carry (YPC) and Yards Per Attempt (YPA), as well as points scored and points allowed. The formula to calculate “Team Strength” is as follows:

(YPC (offense) + YPA (offense) + Avg Pts/Game Scored) – (YPC (defense) + YPA (defense)+ Avg Pts/Game Allowed)

The formula has proven to be a pretty accurate predictor of success. Even in the first week of the 2008 season, 5 of the top 10 ranked teams were playoff bound. As with any statistic, it becomes more meaningful as the sample size grows. Usually, these become most meaningful after Week 3. In 2007, 9 of the top 10 ranked teams were playoff teams, with the lowest ranked playoff team coming in at #15. In 2008, 8 the top 10 were playoff teams, with Arizona being the lowest ranked playoff team at #19. I’m not sure any formula could have predicated their run.

If you’d like to see how teams rankings changed from 2007 to 2008, you can read more here.

I have switched over to sharing the rankings via Google Docs, so hopefully, they are still readable.

Without further delay, here are the first rankings of 2009. It is the first time that the Seahawks have ever been #1. Bask in the glow!

Founder, Editor & Lead Writer
  1. Valid point. SOS is not part of the formula. I have not found a great way to factor that in. Suggestions?

    Keep in mind, I want this to be quantitative, so ideally I'd want to factor in the calculated team strength into the SOS formula.

    All that said, the results speak for themselves. This formula has been reasonably predictive the last two seasons.

  2. I would be curious to see if this could be made more accurate/robust with the addition of some sort of special teams stats – avg kick return, net avg punt return (with something to take into account field position?) – and/or something to account for turnovers (either the value of a turnover from one of the stats websites or just avg. or the +/- differential).

    That way, it might at least somewhat take into account the quality of the opposition – i.e., committing three turnovers against the Rams at home might not be as detrimental as three turnovers against the Steelers in Pittsburgh.

  3. Oh and great start on taking basic figures and having some correlative success. Rankings are certainly one of the more interesting problems for statisticians and the general public alike. See BCS standings.

    Accuracy might not be best judged by playoff appearances, considering the 9-7 Cardinals can go to the playoffs as NFC West champs, and leave another division's second or third team out of the running. If you are trying to judge accuracy by playoff appearances, then it would seem that the model would eventually have to incorporate either something to account for divisions, but I don't think that is what "power rankings" are for, right? If the purpose is an overall ranking, then you wouldn't want to incorporate that.

    Also, I would be interested in a number to account for a Win vs. a Loss – thereby putting the "most important" outcome into consideration with the model – i.e. the Pats in your model would be over the Bills because they did win (as much as I hated to see it).

    In that vein – from an optimization perspective, it seems that, after the first week, you could weight each team's points with the rank (x/32) and or the difference between ranks to account for the strength and relative strength of the opponent.

    I am not sure how well all of this would work out in practice, and am just brainstorming ideas.

    It is fun to think about, though! Thanks for sharing!

  4. Chris, thanks for the ideas! I'm a little slow on the uptake, would you mind showing me what the formula would look like that you are proposing?

    I'm not sure I agree on the special teams part, but weighting the team's points sounded interesting.

    I tend to fall on the side that Devin is promoting, which is that it's easy to overthink these things. Point differential alone tells a pretty significant story. I add in passing and rushing efficiency to give a little greater perspective on how a team is getting to those points (scored for or against).

    The best measure of this (or other) systems would be how often a team ranked ahead of another wins in head-to-head matchups. As a full-time employee and father of two, I simply don't have the time to track that.

Comments are closed.