‘Go For It!’ — Why NFL teams are going-for-it on 4th downs more than ever
As I was watching football games throughout the 2021 NFL regular season, it felt like teams were going-for-it on 4th down at a higher rate. After the season ended, I came across a Twitter thread from Michael Lopez, the Director of Football Data and Analytics for the NFL, which confirmed my prior.
In the same thread, Lopez posts a plot that displays the go-for-it rates in 4th-and-short situations for the past 10 seasons.
This sparked my interest in digging deeper into the numbers as I had numerous questions. What is the go-for-it rate for 4th-and-short opportunities before the 2011 season? How effective is going-for-it in 4th-and-short situations? What teams are going-for-it on 4th-and-short the most?
In order to answer these questions, I needed data. Luckily, a great team of smart people created the nflfastR package which allows for efficient scraping of NFL play-by-play data from the 1999 season. Since I had no experience in working with nflfastR, I found an awesome Python tutorial made by @DeryckG_ that provided great help throughout this project.
The 4th-and-short Go-For-It Rate
The first thing that I wanted to do was identify the trend of go-for-it rates in seasons prior to 2011 by re-creating the plot that Michael Lopez tweeted. I decided to use 15 seasons of data starting from the 2007 season. I did this by isolating the data with only the 4th-and-1, 4th-and-2, and 4th-and-3 plays. I then grouped by season and found the go-for-it rate by dividing the number of passes and runs over the total number of fourth-and-short plays.
In addition, I also filtered the data using win probability which denotes a team’s current game win probability for any given play in order to filter out the plays where the game isn’t competitive at that point. I removed the 4th-and-short plays where a team’s win probability was less than 20 % and greater than 80 %. Here is the result:
The data suggests that in the five seasons prior to the 2011 season, 4th-and-2 and 4th-and-3 rates were stable and there was a general decline in 4th-and-1 rates. However, we see that there was a dramatic increase in teams going-for-it on 4th-and-1 in 2021 at almost twice the rate compared to 2017. We also see that more teams went-for-it on 4th-and-2 in 2021 than teams went-for-it in 2011 on 4th-and-2. This shows that teams are being more aggressive in trying to keep the ball to maximize the number of points they can score in a given drive.
How effective is going-for-it in 4th-and-short situations?
The next thing that I wanted to look at was if over these past 15 seasons, was the league having success in going-for-it in these 4th-and-short situations? In order to calculate this, I used WPA (Win Probability Added) and Expected Points Added (EPA). WPA describes the amount in which a team’s win probability in a game increases or decreases based on a given play.
EPA is based on EP (Expected Points) which attempts to quantify the expected points of scoring based on a given down and distance since not all yards gained are the same. EPA measures how well a team performs relative to the expectation. From a The 33rd Team article, “say the Chiefs start with the ball first-and-10 from their own 25-yard line, where its expected points would be about 1.06. If Patrick Mahomes throws a 15-yard completion, making it first-and-10 on the KC 40-yard line, where the expected points is now 1.88, the EPA of that play would be 1.88–1.06 or 0.82”.
Given these two metrics, I took the median WPA and EPA per season of 4th-and-short plays and grouped by the different play types of pass, run, field goal, and punt. Here are the plots:
We see that kicking a field goal and punting in 4th-and-short situations consistently produces a negative average WPA with kicking a field goal producing a slightly positive EPA. However, running the football in such situations consistently produces on average, an increase of about 5% in win probability and around 1.8 expected points. While the result in passing the football in these situations fluctuate, it has shown to produce positive WPA and EPA in the past 4 seasons which is significantly better than punting or kicking a field goal given where a team is on the field. This illustrates that in 4th-and-short situations, a team is incentivized to go-for-it the vast majority of the time.
We saw an example of this in the final game of the 2021 regular season with the Los Angeles Chargers and the Las Vegas Raiders. The Chargers had a 4th-and-1 at their own 18-yard line and they decided to go-for-it. The Chargers failed to convert as the runner was short of the line to gain. However, while many will disagree with the decision to go-for-it, the thought process was correct as it was justified by many different models including this one from Next Gen Stats:
How often did teams go-for-it this past season?
The last thing that I wanted to do was find out how frequently teams went for it in 2021 on 4th-and-short plays. To get this information, I summed the number of times a team passed or ran in 4th-and-short situations and divided it by the number of plays a team ran in 4th-and-short situations. I then followed the nflfastR Python tutorial referenced earlier to get each team’s logo and color information to make a bar chart. Here is the result:
As seen from the chart, the Indianapolis Colts went-for-it at the highest rate at about 68%. This is not surprising as Colts head coach Frank Reich has been candid about being aggressive on 4th downs. On the other end, the Seattle Seahawks went-for-it at the lowest rate at about 15%. This shouldn’t be seen as a surprise either as Seahawks head coach Pete Carroll is widely considered as one of the most conservative head coaches in the NFL when it comes to 4th downs.
The emergence of analytics in the NFL
While teams are going-for-it more than ever, there are still teams that are far too conservative and costing their team valuable opportunities to increase their chances of scoring given what the data suggests.
In fact, there are many models that predict if teams are making the right decisions on 4th downs. One that has caught my eye is made by Ben Baldwin from The Athletic. Baldwin wrote a fantastic article on how he created a 4th down model that includes a Twitter bot which tweets out the optimal decision and result of every 4th down that occurs in real time! Here is a tweet from the yesterday’s playoff game between the Cincinnati Bengals and the Tennessee Titans on the Titans’ last 4th down attempt:
The increasing trend of teams going-for-it on 4th down coincides with more teams embracing analytics by hiring talented individuals to operate their football research and analytics departments. Colts head coach Frank Reich, who I mentioned earlier, credits John Park, the team’s Director of Football Research, with helping him with in-game decision-making on 4th downs. Like Reich, Chargers head coach Brandon Staley and Ravens head coach John Harbaugh also praise their analytics departments and have confidently spoken about being more open to going-for-it on 4th downs. The question is, will more head coaches do the same in the future?
A link to the code for this project can be found here.
