NFL team efficiency rankings are back for 2008. The ratings are listed below in terms of generic win probability. The GWP is the probability a team would beat the league average team at a neutral site. Each team's opponent's average GWP is also listed, which can be considered to-date strength of schedule, and the ratings include adjustments for opponent strength.
GWP is based on a logistic regression model applied to data through week 3. The model is based on offensive and defensive passing and running efficiency, offensive turnover rates, and team penalty rates. A full explanation of the methodology can be found here. This year, however, I've made one important change based on research that strongly indicates that defensive interception rates are highly random and not consistent throughout the year. Accordingly, I've removed them from the model. Removing defensive interception rates from last year's prediction model did not harm its predictive ability by a single game.
There will be some surprises in the rankings, so take them with a heavy grain of salt. The results are so nutty, I considered not posting them. But the whole point of this site is to let the numbers speak for themselves, so here they are. This is the time of year that makes the least intuitive sense. Most fans, including myself, are wrapped up in how good various teams "should be" or "were last year." But those notions are based on old information mixed with media hype. With only 3 games of data, there's not a lot to go on. But at this point last year, 10 of the top 11 teams in GWP rankings went on to make the playoffs.
After week 3, the top team is 2-1 Washington, which stands out by virtue of its zero turnovers--no interceptions and no fumbles, lost or otherwise. Last year's AFC South powerhouses Indianapolis and Jacksonville are ranked 23rd and 24th , below even hapless Cincinnati. The explanation is in their respective strengths of schedule. The AFC North might be the strangest division, with 0-3 CIN leading in efficiency, and 2-0 Baltimore 3rd out of 4 teams. Lots more surprises below.
Click on the table headers to sort.RANK TEAM GWP Opp GWP 1 WAS 0.72 0.61 2 ARI 0.71 0.60 3 SD 0.70 0.50 4 DEN 0.63 0.57 5 SF 0.62 0.45 6 NYG 0.61 0.40 7 BUF 0.60 0.45 8 DAL 0.58 0.44 9 NO 0.57 0.55 10 PHI 0.56 0.47 11 TB 0.55 0.57 12 CAR 0.54 0.59 13 NYJ 0.53 0.55 14 MIA 0.51 0.52 15 ATL 0.51 0.34 16 CHI 0.51 0.47 17 CIN 0.51 0.64 18 MIN 0.51 0.49 19 GB 0.49 0.48 20 OAK 0.49 0.50 21 SEA 0.48 0.51 22 TEN 0.47 0.36 23 IND 0.47 0.48 24 JAX 0.47 0.59 25 PIT 0.42 0.37 26 HOU 0.40 0.58 27 NE 0.38 0.42 28 BAL 0.38 0.29 29 DET 0.37 0.62 30 KC 0.32 0.53 31 STL 0.29 0.58 32 CLE 0.28 0.60
- Home
- team rankings
- Week 3 Efficiency Rankings
Week 3 Efficiency Rankings
By
Brian Burke
published on 9/24/2008
in
team efficiency,
team rankings
Subscribe to:
Post Comments (Atom)
I think you have sample size issues. But yeah, nothing you can do about that.
Any thought in using the last 2 or 3 weeks of the previous season to smooth out these early rankings? It would bias teams that made big changes in the off season (or had their 2-time SB MVP QB lose a leg) but would help with the small sample. The net effect might be a better ranking, who knows.
p.s. I haven't read the full methodology so if you are already doing something like this go ahead and delete this comment.
Good points above. What I actually do is add several weeks of neutral dummy stats to each teams' stats. This way, it regresses extreme performance in the short 3 weeks of the season. For example, WAS's zero interception rate to-date is averaged together with 5 weeks of the league-average int rate. It gives them a reasonable 1.5% int rate for now.
Then, as each additional week of data comes in, I reduce the number of dummy weeks by 1. By week 8, team stats are generally stabilized at near their steady state season-long average. So by then, there are no more dummy weeks. See the graph in this post.
The main purpose is to reduce the extreme over-confidence of the game win-probability estimates. The ranking order of teams isn't affected, but the spread of 'GWP' is narrower (and more realistic) among teams.
There are a number of ways to stabilize the early-season performance. My priority is to make everything as objective and opinion-free as possible. Using last year's stats as a starting point would do that, and make a lot of sense, but as pointed out, it could be dangerous. Look at the Colts and Pats this year, or the Bears and Ravens last year.
Using a subjective pre-season rating is another possibility, but it's not purely objective. Besides, if you're just looking for some guy's opinion, there are 1,000 places to get that. I'm trying to be the antidote to that stuff.
So yes, beware small sample sizes. But the sample of pass attempts and run attempts is much larger than the number of games in the season so far.
Can you also post the individual components of your model for each team, so we can sort on each one to see who ranks were. You did something like this last year occasionally and it was very informative. Thanks.
Why did you remove the defensive interception rates in your regression? I thought the variable was significant. Did you apply the regression on the last 5 years, or did you keep the coefficients of last year regression?
As usual very good information on your blog.
Now that you have gotten rid of defensive interception rate, is this season's model simply last season's equation with that variable removed? Or did taking out defensive interception rate cause changes to the coefficients for the other variables?
Now that you have gotten rid of defensive interception rate, is this season's model simply last season's equation with that variable removed? Or did taking out defensive interception rate cause changes to the coefficients for the other variables?
SPS-See these articles:
Turnovers and Expected Wins ,
Signal vs Noise and
Explanation vs Prediction
The short answer is that, yes, def ints are very significant in explaining past wins. But they are so random and inconsistent throughout the year that they are not predictive. We all know how good teams were. We can just look at their record or point differentials. My goal is to gauge how good teams really are and will be in the future.
For now, only because I'm short on time, I left the coefficients alone and assigned all the teams the average def int rate. Next week I'll have new coefficients (weights for each efficiency stat), to reflect the removal of def ints.
Thanks for the great questions. Sorry I didn't explain all this in the post, but I didn't want it to become an eye-chart of technical detail.
Also, yes, I can post the efficiency stats that are the components of the model. If I don't have time, I might have to wait until next week.
"Using a subjective pre-season rating is another possibility, but it's not purely objective. Besides, if you're just looking for some guy's opinion, there are 1,000 places to get that. I'm trying to be the antidote to that stuff."
What about the "The Wisdom of Crowds?"
Very true! How about a "power ranking" aggregator? Kind of like realclearpolitics.com, but for football?
I'm a bit cynical about the practicality of the wisdom of crowds here; the existence of systematic errors, as opposed to random errors, will bias the outcome of a survey or market, as you're probably aware. Network effects and confirmation biases might be in play in power rankings, not to mention that only in gambling markets is there the necessary incentive to be right, whereas "some guy's opinion" is usually just that. So I'll stick with objectivity, and I really appreciate your work.
The existence of systematic errors (biases in telephone polling techniques, standardized weights, and again, network effects) is a big problem for realclearpolitics.com. They may be better than any one poll by CBS or NBC, but still systematically inaccurate.