Creating Yards Over Expected and Using it to Measure Offenses in the NFL
The current state of the NFL is all about high powered offenses, big explosive plays, and playing aggressively and efficiently. We see players like Mahomes and Josh Allen stretch the field and push the ball every chance they get. But how do we know which offenses are the best and most effective and which ones absolutely suck? My goal of this project was to create a stat of how many more yards a team gains than they are expected to and use that to analyze the league’s landscape at this point in time.
How does the stat work?
For this stat I am taking into account down, yards to go, yardline, time left in the game, and score differential as my training and testing features. I am plugging these features into a simple Deep Learning regression model where it will give me an expected yards stat. The difference between total yards gained and total expected yards is the final stat Yards Over Expected. I am using data from the past 3 NFL seasons to train my model, and am sourcing my data from the nflfastr dataset.
How do we know this stat is accurate?
If you look at the graph below we can see that YOE actually has an incredibly strong correlation to EPA. EPA is one of the best success metrics in the world of football statistics, and it has a strong correlation to wins — check out some of my other articles for more in depth explanations of EPA. The graph here displays that the higher your YOE, the higher your EPA, proving that the stat is very good for measuring how good a team is.
So… who is the best offense in the league?
According to the stat, as of October 30(week 8), the best offense in the NFL is the Los Angeles Rams. We knew something along these lines was going to happen when the Rams pooled all their chips to make a move to get Matthew Stafford. Led by one of the best offensive minded coaches in the league and a top 10 quarterback this isn’t really much of a surprise.
Baltimore, Tampa, and Dallas have all been showing out as well, and even though their record doesn’t show it, the Chiefs have been cooking as well. We can see here that the Bears, Jets, and Texans are all bottom in the league, and it makes sense. Starting a rookie QB right away seems to have a bad effect on the offenses and plugging them in right away isn’t the best thing to do. We can also see how the Sam Darnold trade didn’t really make the Panthers' offense much better, and they’re actually bottom 5 in the NFL. The Titans were widely regarded as one of the best offenses in the league and due to their slow start they’re barely in the top 20, but they have started to find their footing in the past couple games. The Bengals have been one of the most surprising teams in the league this year, and we can see how good they’ve been, with them being top 10 in the metric. Even though the Cardinals have the best record in the league with an elite offense we don’t see them top 5 in the stat, showing how balanced their team actually is.
What is also interesting with this is to see how much parity there actually is in the NFL this year. All the teams in the top 9 are extremely close in this metric, barring the Rams, and this year’s postseason should be incredibly close.
This metric can also be used to find trends within certain players, as well as coaches. We can also filter out individual players’ YOE to evaluate how much they bring to an offense and whether they are helping or hurting their team.
This stat helps us understand how the league is playing out, it helps us measure how powerful each powerhouse actually is, as well as looking at things like how rookie quarterbacks effect an offense, and how trades have changed offense for the better or worse. This may be the most competitive the NFL has ever been, and we can’t miss history in action
Link to the code: https://github.com/nickraisgit/NishankMLRepo/blob/main/AvgYardsOverExpected.ipynb
Contact: rai.r.nick@gmail.com
Data from: nflfastr and Nishank Raisinghani
