September 6, 2018 - by David Hess
Here are the official TeamRankings 2018 NFL preseason rankings and ratings. Further explanation of our preseason ratings methodology and tips for interpreting the data follow below the table.
Our 2018 NFL preseason rankings are almost entirely data-driven. We’ve used team data from past seasons to find which descriptive statistics have correlated strongly with high end-of-season power ratings. We then used those stats to create a model that predicts a team’s power rating.
Here are the key components of the model, in order of importance.
Team Quality Baseline. This is based on a weighted average of end-of-season team power ratings from the last few years. The most recent season gets by far the strongest weight.
The highest and lowest Team Quality Baselines this season:
QB Changes. We make simple projections for the QB Rating of all the starting NFL quarterbacks this season. We then compare that to the actual QB rating for each team from the past several years (with a weighted average that mimics the Team Quality Baseline weighting).
The biggest positive and negative QB Change adjustments this season:
Turnover Luck. This one may be familiar to those of you who like to read about football analytics. While he didn’t come up with the idea, Bill Barnwell has written extensively about how teams with good turnover margins tend to get worse the following year, and vice versa. We go deeper than that, and break turnovers down by type.
What we’ve found is that interceptions thrown are mostly a QB skill, and we barely adjust for those at all. On the other hand, opponent fumbles lost is almost entirely luck (after accounting for forcing fumbles, which does involve some repeatable skill), and we adjust very strongly for those. Interceptions caught and own fumbles lost are between those two extremes.
The biggest positive and negative adjustments for past-season Turnover Luck:
Draft Score. We assign “draft points” to every pick in the first 3 rounds of the NFL draft, from 5 points for the top pick to 1 point for a 3rd-round pick. Our research has shown that teams with high Draft Scores from 3 and 4 drafts ago tend to improve the most, probably due to having a lot of improving young players.
This year the largest positive and negative Draft Score adjustments belong to:
After accounting for all these components, we have a power rating that represents how many points above or below average we think a team is. A rating of 0.0 represents a “perfectly average” team.
Once we generate initial 2018 NFL preseason rankings & ratings, we then check them against the betting markets and other preseason ratings.
If our ranking for a team seems severely out of whack with those other sources, we’ll investigate. We check to see if there’s some factor not taken into account by our model, that the betting market is picking up on. Most commonly, this might involve a coaching change, or some major defensive free agent upgrades. If we find something we think warrants an adjustment, we’ll tweak our rating toward the consensus. Though only rarely will we adjust it all the way to match the consensus.
It’s worth noting that these preseason ratings also drive our NFL season projections — at least before the season begins. As the 2018 season progresses, the impact of these preseason ratings will gradually fade, and actual game results will play a larger role in determining our team power ratings (which continue to drive the season projections).
Dear Hardcore Fan Of Team X, before you get angry that our models are obviously biased against your favorite team, please keep two things in mind:
Printed from TeamRankings.com - © 2005-2018 Team Rankings, LLC. All Rights Reserved.