For over a decade, Football Outsiders has been publishing its DVOA grades. Last week, Andreas Shepard, a loyal reader of both FO and this site, came up with estimated DVOA ratings going back to 1950. You can read the fine print on how he derived the formula in the Methodology section at the end of this post. Andreas did an excellent job looking at some of the best and worst teams in many different DVOA categories, so you should give his article a read. But to me, at least, the real value of team ratings for over 1600 teams in 8 different categories is as a reference piece. And since the tables I create here are both sortable and easily searchable, I’ve worked with Andreas to present the team ratings in a way to make life easy for the reader. Consider these like an encyclopedia for team ratings, available for you to find the team you’re interested in whenever you like.
Andreas created estimated DVOA ratings for each year from 1950 to 2013 (remember, Football Outsiders has posted actual DVOA ratings published for each year from 1989 to 2013, but I am providing the estimates for each year.) The table below shows all 1638 teams from 1950 to 2013; here’s how to read the table below, which is sorted from best to worst in Total DVOA. The Packers (you can click the link to see Green Bay’s PFR page that year) in 1962, playing in the NFL, rank as the top team in estimated DVOA. That year, Green Bay went 13-1 with a winning percentage of 0.929 and a points differential of 19.1 points per game.
The Packers had an Offensive Pass DVOA grade of 18.9% and an Offensive Run DVOA grade of 23.0%; all DVOA ratings are centered around 0%, so this shows how the Packers were well above average in both offensive measures. For defensive ratings, negative grades are better, and the Packers have an incredible -33.2% estimated Defensive Pass DVOA grade, along with a -8.6% estimated Defensive Run DVOA grade. The Packers’ Total Offensive DVOA grade was 21.4%, the Total Defensive DVOA grade was -20.7%, and the Special Teams DVOA grade was 5.8%. Finally, the Packers have a total estimated DVOA of 47.9%, the best since 1950. If you type “gnb” into the table below, you will see all Packers teams. You can type in any team’s code to see just their teams, or sort the table by any of the categories available.
There’s way too much information here to do it justice with a writeup, so I’ll only make one small note. The best pass offense belongs to Otto Graham’s Browns, although that’s not too surprising. He didn’t have the gross numbers of say, Peyton Manning‘s 2004 Colts, but in 1953, Graham produced the single best yards/attempt average since 1950.
The next table takes all that same information but provides each team’s rank in that league year. So in 1962, the Packers led the NFL in well, just about every category except for Offensive Pass DVOA, Defensive Run DVOA, and Total Defense. If you type “1962 NFL” into the search box and then sort the appropriate columns, you will find that the Giants (led by Y.A. Tittle) led the NFL in estimated Offensive Pass DVOA, while the Lions — who outscored the Packers in two games, and had a dominant defense featuring Roger Brown, Alex Karras, Joe Schmidt, Wayne Walker, Night Train Lane, Dick LeBeau, and Yale Lary — finished first in the NFL in both estimated rush defense and estimated total defense. Here’s another neat fact: the 1972 Dolphins are the only NFL team to rank #1 in both estimated total offense and estimated total defense.
Finally, I thought it would be useful to present the team rankings over the entire 63-year period. So each team gets a rating in each category from 1 to 1638. Remember, the tables are fully sortable and searchable, so here are the number one teams in each category:
Offensive Pass – 1953 Cleveland Browns
Offensive Run – 2011 Carolina Panthers
Defensive Pass – 1991 Philadelphia Eagles
Defensive Run – 1991 Philadelphia Eagles (!)
Total Offense – 1953 Cleveland Browns
Total Defense – 1991 Philadelphia Eagles
Special Teams – 2009 Cleveland Browns
Total – 1962 Green Bay Packers
Remember, we have actual DVOA ratings for the last 25 years, so there’s a limit to the value of some of these estimates. They’re very close, of course, but I’ll note that the 1991 Eagles rank 2nd in both actual pass defense (to the 2002 Bucs) and actual run defense (to the 2000 Ravens), while the 2009 Browns rank a bit lower when it comes to actual DVOA ratings. In any event, I want to thank Andreas for coming up with these ratings, and allowing me to present them to you.
Methodology
The following was explained to me by Andreas via email:
My overall goal was to approximate DVOA as closely as possible, not to come up with my own assessment of which team was the best and worst. For that reason, I tried to mirror DVOA’s structure as closely as possible. Since DVOA compares performance to average on a per play basis, I did the same. I only considered teams from 1950 to present. That was the year of the NFL-AAFC merger, which also marked a) the beginning of more complete and reliable statistics and b) the approximate end of the era when teams would start up and fold within a few seasons.
For each team season, I collected as many box score statistics as I could find. These are all season-long statistics, and for 1987 they include the 3 replacement games as well as the 12 non-replacement games. The full list is:
Passing offense and defense: completions, attempts, gross yards, TD, INT, sacks (estimated before 1969), sack yards, first downs
Rushing offense and defense: attempts, yards, TD, first downs
Overall offense and defense: total fumbles (lost and recovered)
Special teams (for and against): FGM, FGA, XPM, XPA, kick returns, KR yards, KR TDs, punts, punt yards, punt touchbacks (1976+ only), punts blocked, punts inside the 20 (1976+ only), punt returns, PR yards, PR TDsI then divided passing statistics by (pass attempts + sacks) and rushing statistics by rushing attempts, to find averages per play. I then divided each teams per play stats by the weighted average for the league that season to find how much better or worse per play they were than the average. I used each season as a separate baseline, because the baselines for DVOA are also different for each season. For the pre-merger AFL and NFL seasons, I set the baseline using the weighted averages for the whole league. In other words, I did not give any penalty to the AFL teams for playing in a worse league. Ideally I would have liked to, but I haven’t yet been able to find a methodology I was satisfied with.
To convert these to indexed per-play statistics into DVOA, I ran several linear regressions on actual VOA using teams from 1989-2012. I used VOA, the version of DVOA not adjusted for opponent, because the indexed box score statistics are also not adjusted for opponent. The resulting equations were:
Passing offense/defense VOA = -84.22 + 56.19 x (indexed NY/pass) + 21.64 x (indexed completions/pass) + -6.45 x (indexed sacks/pass) + -20.73 x (indexed int/pass) + 36.27 x (indexed first downs/pass) + -7.48 x (indexed total fumbles/offensive play). R^2 = 0.902
Rushing offense/defense VOA = -73.01 + 44.51 x (indexed yards/rush) + 7.21 x (indexed TD/rush) + 26.80 x (indexed first downs/rush) + -7.29 x (indexed total fumbles/offensive play). R^2 = 0.751
FG/XP rating = 0.46 + 0.20 x (XPM) + -0.24 x (XPA) + 2.08 x (FGM) – 1.60 x (FGA). R^2 = 0.751
Kickoff rating = 5.14 + 0.32 x (Opp kick returns) + -0.02 x (Opp KR yards) + -4.71 x (Opp KR TD). R^2 = 0.314
Kick return rating = 1.57 + -0.62 x (kick returns) + 0.03 x (KR yards) + 4.73 x (KR TDs). R^2 = 0.567
Punting rating (1976-present) = -3.32 + -0.34 x (punts) + 0.01 x (punt yards) + -0.22 x (touchbacks) + -3.18 (punts blocked) + 0.22 x (punts inside the 20) + 0.33 x (opponent punt returns) + -0.05 x (opponent PR yards) + -2.55 x (opp PR TDs). R^2 = 0.602
Punting rating (1950-1975) = -3.13 + -0.34 x (punts) + 0.01 x (punt yards) + -3.26 x (punts blocked) + 0.31 x (opponent punt returns) + -0.06 x (opponent PR yards) + -2.32 x (opp PR TDs). R^2 = 0.579
Punt return rating = 1.14 + -0.04 x (opponent punts) + -0.45 x (punt returns) + 0.05 x (PR yards) + 2.43 x (PR TDs). R^2 = 0.732Plugging each team’s stats into this equation gives estimated VOA for passing and rushing offense and defense, and for each of the major special teams disciplines. To convert these into final DVOA, I took several steps. First, I scaled the VOA estimates so that the standard deviation of the estimated VOAs matched the actual standard deviation for VOA. This is necessary because a linear regression with an R^2 of less than 1 will always give estimates that are less spread out than the actual data. This made the best teams’ ratings 10-20% better, and the worst teams 10-20% worse. Next, I added in opponent adjustments. For each team, I took an average of their opponents estimated VOA for passing and rushing offense and defense, weighted by the number of games they played. I then subtracted the opposing averages from the team’s estimated VOA (opponent passing defense from passing offense, opponent rushing offense from rushing defense) to get estimated VOAs. Like in real DVOA, special teams did not get an opponent adjustment.
The final step is to summarize these individual DVOAs into composite ratings for offense, defense, and special teams. Special teams is just a simple average of the ratings in the 5 disciplines, minus the average estimate DVOA for that season (to account for the secular improvement in special teams statistics over time). Defense is a weighted average of passing and rushing defense (weighted by passing and rushing plays faced). Offense starts as a similar weighted average, but then has one final adjustment. There are a number of plays, like false starts and aborted snaps, which count neither for the rushing or passing offense (or the opposing defense) but still factor into total offense. For 1989-2013, I found the actual amount of this factor (generally -4 to -6% DVOA) and applied it to each teams offense. For teams before 1989, I used the average factor for the years I did have (almost exactly -5%).
With that final adjustment made, all that’s needed to calculate total DVOA is to add Offensive DVOA + (-Defensive DVOA) + Special Teams DVOA (reversing defense because better defense is negative). The end result has a correlation to actual DVOA of 0.958.
I’m sure Andreas will be checking the comments to this post, so leave any kind words or questions for him here.
