Analytics in football have not progressed as far as analytics in basketball. The NBA’s primary advanced metric that people use to assess overall team performance is net rating. More specifically, net rating is found by subtracting a team’s defensive rating, the points allowed per 100 possessions, from their offensive rating, the points scored per 100 possessions. However, the NFL currently does not have any stat like this. Most people still use points per game or yards per game to rank the offenses and defenses in the league. This method of ranking comes with many flaws. Primarily, it does not consider that some teams may have fewer plays per game or even drives per game than other teams, and they also do not account for starting field position or strength of schedule. Even more baffling, when teams score defensive touchdowns, they count as points scored, which is used to evaluate offense.
With this project, I seek to create an offensive and defensive rating for the NFL to be used as alternatives for evaluating offense and defense to simpler stats like points per game or yards per game.
The goal was to make this stat closely resemble offensive and defensive rating in the NBA, and make any other improvements necessary to improve its overall descriptiveness. Since offensive rating in the NBA is found by looking at each possession separately, I did a similar thing for the NFL by using drives. Possessions in the NBA and drives in the NFL are similar since both provide an opportunity to score and usually end with the ball being in the hands of the other team. Therefore, the first step was to calculate the points per drive for each team. However, this still does not take field position or strength of schedule into consideration.
In order to see the effect of field position on points per drive, I plotted the average points per drive for every starting field position this season, excluding drives consisting of only QB kneels. Then, I used a cubic regression to give every starting field position an expected points per drive value. The graph and regression output are shown below. From this, we can see that not all possessions all equal. For example, a possession that start at a team’s own 25 yard line is expected to score about 2 points on average, while a possession starting at the 50 yard line is expected to score about 3 points on average. The values do not move linearly, though, since if a possession starts within the opponent’s 5 yard line, it is expected to score close to 6 points on average.
After running the regression, I applied the results to all the non-QB kneel drives in the 2020 NFL Season. Then, I found the points over expectation for each drive and averaged them for each team’s offense and defense. These values were the raw offensive and defensive efficiencies. The defensive efficiency was manipulated so a positive value is better. After this, the last thing that needed to be accounted for was strength of opponent.
Adjusting for strength of schedule was done by averaging the raw offensive and defensive efficiency faced for each team. Then, the adjusted offensive efficiency was calculated by adding the raw offensive efficiency and the average defensive efficiency faced, and vice versa for adjusted defensive efficiency. This way, offenses that played good defenses were rewarded, while offenses that played bad defenses were penalized (and vice versa for defense). Finally, offensive and defensive rating were found by multiplying the adjusted offensive efficiency and adjusted defensive efficiency by 10, since NFL teams usually have close to 10 possessions per game.
The final values for offensive rating and defensive rating were centered around 0, with offensive rating having a standard deviation around 3.9 and defensive rating having a standard deviation of approximately 3.2. The values had a mean close to 0 because of the step where I took the points above expectation per drive, which means that teams with positive values are above average while teams with negative values were below average. Both distributions were close to being normal distributions, but were very slightly right skewed. The final rankings for the NFL teams in the 2020 season based on offensive rating and defensive rating are shown below. Total rating is offensive rating plus defensive rating.
Analysis of Findings
Using these offensive and defensive ratings, we can see which teams were good in both areas, just one area, or struggled at both.
The teams located in the top left corner of the graph above have good offenses and defenses. This year, only three teams had top 10 offenses and defenses: the Buccaneers, Saints, and Ravens. Some other teams with high total ratings, such as the Bills, Rams, Seahawks, Steelers, and Titans, are very imbalanced. The Bills, Seahawks, and Titans have very good offenses but bad defenses, while the Steelers and Rams excel at defense but not offense. Only three teams have bottom 10 offenses and defenses as well. They are the Jets, Jaguars, and Bengals.
By using the newly created offensive and defensive ratings, we can also see which teams are underrated and overrated by points per game. On offense, the Colts and Steelers are worse than their points per game suggest, while the Giants and Patriots are better. On defense, the Broncos, Vikings, and Jaguars were all better than their points allowed values, but the Patriots and Ravens were worse. The teams that are underrated by points per game are likely hurt by their starting field positions or number of drives per game, while the teams that are overrated by points per game are likely aided by their starting field position and number of drives per game.
Using the total efficiency, the Colts and Ravens were the most overrated teams by points per game. The Colts had a margin of victory of 5.6 when they should’ve had a margin of victory closer to 0.7, and the Ravens had a margin of victory over 10 even though they should’ve had a margin of victory closer to around 7. Meanwhile, the 49ers, Vikings, Broncos, Packers, and Rams all should’ve had larger margins of victory than they did. The Rams and Packers may be sleepers teams in the playoffs because of this, while the Colts and Ravens may have early exits.
Predicting the Wildcard Games
One further application of the NFL offensive and defensive ratings would be using them to give the probability of a team winning against a certain opponent. In order to accomplish this task, I used the data from all 255 non-tied games this season and applied a logistic regression with two inputs: the difference in total rating and a home field advantage variable (equalled 1 for home, -1 for away). The regression output for the log odds are shown below.
Using the total rating to predict games produced very reliable results, shown by the extreme significance of the total rating difference (Diff) variable. However, home field advantage has essentially been diminished this season, as it was relatively insignificant in the model. After applying this model to the six wildcard games coming up this weekend, I found the probability of each team winning their game.
- Bills 73.9%, Colts 26.1%
- Rams 61.2%, Seahawks 38.8%
- Buccaneers 81.9%, Washington 18.1%
- Ravens 65.7%, Titans 34.3%
- Saints 76.3%, Bears 23.7%
- Steelers 71.8%, Browns 28.2%
The wildcard round of the 2021 NFL playoffs features six teams instead of the normal 4 as a result of the NFL’s playoff format change. The Bills-Colts matchup features the number 2 offense, led by Josh Allen, facing off against the number 11 defense, led by Darius Leonard. Then, the Seahawks play the Rams in what could be an upset, as the Rams are the number 5 team in the NFL according to total rating. This matchup also places strength against strength, with the 1st place Rams defense playing the 5th place Seahawks offense. The next matchup is the most lopsided, as the Buccaneers rank 2nd in total rating while Washington ranks 18th, even with their 5th ranked defense.
On Sunday, the 7th ranked Ravens play the 11th ranked Titans. The Titans have the 3rd best offense, led by Derrick Henry, but their defense has struggled, ranking 28th in the NFL. In the late afternoon, the Saints play the Bears. The Bears are the opposite of the Titans as their defense ranks 3rd and their offense ranks 28th, but the Saints have both a top 10 defense and offense. Lastly, the Steelers play the Browns in a top 10 defense vs top 10 offense matchup. However, the Steelers also have a bottom 10 offense and the Browns have a bottom 10 defense.
As for the Super Bowl favorites, the team with the highest total rating is the Green Bay Packers by a pretty large margin of about 2.5 points. The three best teams according to total rating are in the NFC, with a potential divisional round matchup between the 1st place Packers and 2nd place Buccaneers around the corner should the Saints, Seahawks, and Buccaneers win their wildcard games. In the AFC, the Bills actually have the best total rating, but the Chiefs are just 0.08 points behind them. The sleeper candidates of the playoffs are the Buccaneers, Rams, and Ravens since all three are in the top 10 of total rating but only got wildcard spots in the playoffs. The largest bust candidates of the playoffs are Steelers and Seahawks, both having won their division despite not having the division’s best total rating.
My NFL offensive and defensive ratings should be used in place of points per game or yards per game to evaluate NFL teams. They use the team’s average points over expectation based on field position on a per drive basis, which is then adjusted for strength of schedule then multiplied by 10 to give the final ratings. The Packers, Buccaneers, and Saints rank in the top 3 of the 2020 NFL total ratings, with the Packers having the top ranked offense and the Rams having the top ranked defense. The only three teams with a top 10 offense and defense are the Buccaneers, Saints, and Ravens. Based on points per game, the most underrated offense is the Giants, underrated defense is the Broncos, overrated offense is the Colts, and overrated defense is the Patriots. The most underrated teams overall based on margin of victory are the 49ers and Vikings, while the Colts and Ravens are the most overrated. The Super Bowl matchup pitting the team with the team with the best total rating in the NFC against the team with the best total rating in the AFC would be a Packers-Bills Super Bowl. These NFL total ratings are better measures of offensive and defensive success than points per game or yards per game.