Week 3 NFL Spread Projections
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:
Week 2 Best Bets: 4-1 ATS | Overall Best Bets: 8-2
Week 2 Overall: 9-6-1 ATS | Overall Record: 21-10-1 ATS
After two successful weeks for Jon Von Model it’s hard not to get excited. The best bets from the model have Mitch Moss and I in contention for the first three-week contest in the SuperContest, and the overall returns are better than I could have hoped for.
However, I have been tamping down my excitement, because it is an extremely small sample size of success based on an extremely raw statistical model. Still, there is some pride felt on my end.
Week 2 Recap
The biggest edge from last week was Tampa Bay over Chicago. Jon Von Model had projected a 10.89-point victory for the Buccaneers. Thanks to an ill-advised pass on a screen from Justin Fields, Tampa Bay was able to win by 10 points and cover as 2.5-point favorites. That was one of three games that fell right in line with Jon Von Model’s projections.
JVM projected an eight-point win for Kansas City in Jacksonville, and the Chiefs ended up winning by that exact margin. Pittsburgh did the same thing, as it defeated Cleveland 26-22 which is right on the 4.29-point win JVM had projected.
It is obviously a coincidence that these games fell right on the projected margin that JVM had projected, but those results coupled with a winning record on the best bets that this model has turned out has me hopeful that I am on the right track when it comes to building out a successful model.
The Adjustments
With that out of the way, let’s discuss some adjustments for this coming week of football. First, I would like to update you on what’s going on with some of the statistics I was having trouble tracking down.
For those who are just joining the journey, I had run into an obstacle right before the first week of the season when Football Outsiders shut down. Two of the statistics used in the model – net drive success rate and adjusted line yards allowed – were courtesy of Football Outsiders, so not being able to access them was going to be a problem. Well, FTN Fantasy has officially updated its database to include former Football Outsiders metrics, save for drive success rate. As a result, I will be using the RBSDM drive success rate as a substitute.
Now that we have that issue taken care of we can move on to adjusting the model on a week-to-week basis, and this week I was busy doing quite a few things.
One of the biggest adjustments I made this week was to the Los Angeles Rams. When I first ran the matchups for Week 3 the biggest discrepancy between the projections and the betting market was in the Monday night game between Los Angeles and Cincinnati. The first projection Jon Von Model spit out was an 11.36-point win for the Bengals.
That is obviously nowhere near in line with the current spread, and that obviously does not factor in the potential absence of Joe Burrow. So, updated every number for Matthew Stafford and this Rams offense to reflect the team it has been in 2023 as opposed to the one from last season. That immediately shrank the margin to 3.33 points, still in favor of Cincinnati.
As of now, this is the projection we will roll with for this week, because as of now we have no reports that Burrow will miss the game. Should he be ruled out, I will post an updated projection with Jake Browning at quarterback, which will be a challenge, as Browning has one dropback in his NFL career.
The other adjustments I made were to further add to the profiles of the rookie quarterbacks and the quarterbacks which I did not have much data on (Desmond Ridder, Jordan Love). Coming into the season I had given each team that was expected to start a rookie an average signal-caller, or in the case of the players who had been in the league, just their small sample size of statistics were assigned. Now that we have two games of data, I decided to see what inserting each player into the model would do.
Some of the results were extreme.
Seattle was projected to win by 11.92 points after I placed Bryce Young’s current stats into the model. That is a number that is 6.42 points off the current consensus number of 5.5 and it was something I did not feel comfortable rolling with.
Had we stuck with that projection then Jon Von Model would have had three games with edges of more than six points this week, and that is not a sign of an accurate model. Instead, I kept with the current iteration of Young, and as you will see, that line is right in line with the market.
The same phenomenon happened when I plugged in Ridder’s statistics from the first two weeks of the season. JVM projected a two touchdown victory for Detroit when the second-year signal-caller’s numbers were upgraded, and that was already one of the biggest edges of the week. Needless to say, I stuck with the previous iteration of Ridder.
C.J. Stroud’s numbers did not upgrade Houston by any stretch, but his numbers put the projection for the Texans’ game with the Jaguars right on the 9.5-point spread for the game in the market. That is one of the adjustments I decided to keep in the model this week.
Week 3 Projections
Before finding the biggest edges of the week from the model I always cross-reference the projections with the betting market. I want to make sure I do not have eight or nine games with an edge bigger than three points, and once again JVM is right in line with the betting market.
Seven games are within two points of the point spreads on the screen this week, and nine are within three points. There are some big edges on the board, but a vast majority of the contests are tight with the market. It gives me confidence that this is still a usable version of the model.
After that, we find the five biggest edges, and they are as follows:
Chiefs (-12.5)
Ravens (-7.5)
Lions (-3.5)
Commanders (+6.5)
Jets (+2.5)
The first thing that sticks out is the play on Kansas City. For the third week in a row JVM has the Chiefs with one of the biggest edges on the card. In Week 1 we ended up not using Kansas City due to the injury to Travis Kelce and the absence of Chris Jones, but it did come through last week against Jacksonville.
This play is a combination of two things: the model’s love for the Chiefs and its disdain for the Bears.
As I have mentioned many times, this model is based heavily on offensive team statistics and quarterback statistics. Kansas City and Chicago are on the opposite ends of the spectrum when it comes to these, and thus we get a massive edge and the Chiefs on the card once more.
To be perfectly honest, I bristled at first at this projection from the model, but the more I tweaked and tugged the more it became clear that this was the right number. Fields has regressed statistically as a passer, and he was not very good to begin with. This was the worst team in the NFL last season, so why would it be so shocking that JVM views them as just that?
The rest of the best bets I am not too shocked by. If there was one that did stick out it would be the belief that Buffalo is only projected to win by fewer than two points. But, turnover worthy play rate is one of the statistics this is built on, and Josh Allen has a very high percentage.
So, there you have it.
It’s been a busy week of adjustments and tweaks, but I feel like I’ve got a set of projections that I am comfortable with using this weekend. Hopefully the run continues and next week we get to discuss how Mitch Moss and I placed in the first three-week mini-contest of the SuperContest.
I can dream, at least.
