NFL model projections for Week 1
This is part of a recurring series on VSiN.com in which VSiN host Jonathan Von Tobel tracks the progress of a model he created for the NFL season
Top NFL Resources:
- NFL Betting Hub
- NFL 2024 Week-By-Week Schedule
- Parlay Calculator
- NFL Betting Splits
- NFL Betting Odds
If you’re reading this there is a high probability that you have taken in some of the video content we have to offer here at the Vegas Stats & Information Network. Across the countless hours of programming at the network we have some brilliant guests who give up their time to come on VSiN to give their insights on whatever the topic at hand is.
If you watch long enough it won’t take long to hear some iteration of the phrase, “My number for this game is…”
This is what successful bettors do. They spend hours making power ratings and projections, and they then take those projections and bet into the market accordingly. Some do it with good old fashioned pen and paper, and some do it in a more advanced way: modeling.
Many bettors today just need to fire up an Excel spreadsheet, punch in some team names and out comes a projection for that contest which will tell said bettor what, if anything, to bet. Making one of these models takes a lot of time and dedication, and it’s been a task that has fascinated me for many years.
So, I built one.
This summer I dedicated myself to putting together a rudimentary NFL model in order to see what the process is like. It was extremely difficult, but eventually I got to a point where I was able to put together something that resembled a projection system for NFL sides. It is by no means perfect, but it is something to build on and that is exactly what I am going to do in this column all season long.
Here we will go over the model’s projections for the upcoming week of NFL action, discuss some of the flaws that the model has, how to fix them and what adjustments need to be made on a week-to-week basis.
We will also have some fun.
Mitch Moss, host of Follow the Money here on the network, and I have decided to buy an entry in one of the football contests here in town for the sake of tracking the model’s biggest weekly edges. Every week we will use the five best bets the model has as our contest card, and we will track the results throughout the season here in this column.
Before we get into how I put the model together and what the projections are for Week 1 as of Tuesday, Aug. 29 I do want to make something clear about this endeavor: This is a learning experience.
The point of this column and this undertaking is not to give out guaranteed locks and winners. It’s to learn. It’s to show the basics of building out a model like this, and what the process of testing that model is like.
To be fully transparent, I do not believe this model will be successful from an ATS standpoint.
It is a very basic model which will need countless adjustments as the season goes along, but I want to take those lumps in a very transparent way. If anyone reads this column weekly and comes away with some ideas on how to build their own model in a better way then this column did its job.
Also, do not be afraid to reach out on X with any tips on how to make this model better throughout the season. I am open to any and all constructive criticism during this journey.
The Model
So, what exactly did I build? Well, I’m not entirely sure myself, but I can walk you through the process.
Simply put, my model is based on regression analysis. Regression analysis, as I understand it, is a statistical method that tests the relationship between a dependent variable against independent variables in order to predict and forecast. In the model my dependent variable is the home team margin of victory and the independent variables are a myriad of statistics I have tested to find as strong a correlation as possible.
After testing out some different statistics the ones that have turned out the highest correlation to home margin of victory are as follows:
Drive Success Rate (Football Outsiders)
Adjusted Line Yards Allowed (Football Outsiders)
Quarterback Turnover Worthy Play Rate (Pro Football Focus)
Quarterback Passing Grade (Pro Football Focus)
Net Expected Points Added (NFELO)
Offensive Expected Points Added (RBDSM)
Offensive Dropback Success Rate (RBDSM)
Pressure Rate (Pro Football Reference)
Running regression tests on Excel with the home team margin of victory last season against these eight metrics brought back the highest correlation thus far, and as of now this is the model I am rolling with.
The Projections
That is a basic summary of what this model is based on, so let’s get to the good stuff. These are the projections the model, which I will lovingly refer to as Jon Von Model, has for Week 1:
The very right column is the projected margin of victory for the home team, so any negative number would be a projected win for the road team. For example, the Jaguars are projected to win by 8.4 points in Indianapolis, according to Jon Von Model.
As of this week, the largest edges would be Packers (+2), Bills (-2), Jaguars (-4), Chiefs (-6.5) and Eagles (-4.5), and the only real disagreement I would have with the model is the edge on Buffalo.
The model clearly views New York as some version of the team it was last season, and it’s not hard to see why. While I do have Aaron Rodgers plugged in as the quarterback for the Jets, the model is still using the team’s drive success rate, net EPA and offensive EPA statistics from last season. That is clearly a flaw with the model and one I will be working to fix prior to the start of the NFL season.
So, there it is. An introduction to what will be a fun journey throughout the NFL season that I hope you will follow. Next week I’ll work toward some adjustments on teams like the Jets, as I attempt to get the model to have a more accurate read on the changes the team made in the offseason.
