NFL: Week 9 model best bets and projections from Jonathan Von Tobel

NFL Week 9 Best Bets & Model 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

Week 8 Best Bets: 1-4 | Overall Best Bets 21-19
Week 8 Overall: 5-11 | Overall Record: 67-53-2 

Week 8 Recap

Last week was a doozy.

I decided that we would test out the projections from Jon Von Model based on purely 2023 statistics. The result was a 1-4 SuperContest card and a 5-11 record on overall edges; obviously the worst week of the season.

Initially I wanted to draw a line through last week and press on with the 2023 version of the model. Why not? One of the losses in the contest card was the Ravens who allowed the Cardinals to get inside the backdoor on one of the oddest beats of the season. The other was the Browns who were three yards away from a potential outright win, but instead P.J. Walker’s pass bounced off a helmet and Cleveland ended up losing five plays later.

What drew me toward rethinking this whole process was the record on overall edges. Had the model performed better in the other games then it would have given me more confidence in sticking with that version of the model. Instead, a 5-11 record on overall edges – 4-7 in games not involving the contest card – got me to take a step back.

Week 9 Adjustments

This week I decided to go back to the version of Jon Von Model we had been using, but went through some extensive maintenance. The first step was to make sure that every team had the proper quarterback plugged in.

The Raiders, Vikings, Giants, Falcons and Cardinals all have changes at the quarterback position this week. For teams like the Raiders and Giants the process is as simple as plugging in the statistics for both Aidan O’Connell and Daniel Jones. The Falcons were simple as well, as Taylor Heinicke has extensive time as a starting quarterback in the NFL, so his numbers were plugged in as well. For him, I used the best year of his career in Washington. Some might believe that would overrate him as we do not know how he will perform in Atlanta, but his numbers were extremely poor with the Commanders, so I did not believe that would be an issue.

There were bigger challenges as far as Minnesota and Arizona were concerned.

The Vikings are going to start rookie Jaren Hall. Hall had only five dropbacks in the game against Green Bay, and in 64 dropbacks in the preseason he graded out very poorly in the metrics that I use in the model. My process for this at the beginning of the season was to essentially plug in the statistics for an average rookie quarterback, and that is what I did for Hall. The numbers are very below average, but it makes sense for an unknown commodity like Hall.

I did the same for Arizona as well, but the problem is that the door is still open for Kyler Murray to start over Clayton Tune this week. Adam Schefter has reported that it will be Tune under center, so it seems that the choice to “start” Tune on my end is the right one.

When it came to the “extensive maintenance” that I underwent this week, it was about making sure the model did not have skewed views of certain teams.

For example, the model was too high on the Los Angeles Rams. Initial projections had Los Angeles as 6.73-point favorites in this game against Green Bay. I wanted to make sure that there were no statistical outliers weighting the matchup too heavily in favor of the Rams, and I found out that there was.

Los Angeles’ net EPA and net drive success rate had not been properly updated on my end. The numbers that were being used had driven the Rams’ power rating through the roof. A simple adjustment to those statistics to reflect where Los Angeles currently is made all the difference in the world, and brought the number back to within striking distance of the market.

The other big adjustment that needed to be made – and why using the full 2023 version of the model was foolish in hindsight – was to Cincinnati. The Bengals have won and covered three straight games, and Joe Burrow’s play has reflected that of his play last season. Using only 2023 statistics makes the Bengals look much more formidable. So, I decided to essentially plug Cincinnati’s offense from last season into the model. The projection went from Bills -5.32 to a near pick, with a slight edge to Buffalo by 0.52 points.

Week 9 Projections

The projections this week look much better than they did last week. The average edge last week was 5.4 points, but this week the average edge is 1.53 points. That is a number that is much easier to swallow. It makes the margins tighter, but we do not have any edges of over 20 points like we did last week when we went with purely 2023 statistics.

So, a look at the projections would give us the following as our biggest edges of the week:

Titans (+3.67)
Panthers (+3.16)
Bills (+2.52)
Commanders (+2.20)
Dolphins (+1.8)
Seahawks (+1.69)

You will notice that I have included six teams here as part of the discussion. I am doing that because Tennessee will not be part of the contest card this week despite being the biggest edge of the week in terms of pure points, because I do not agree with the assessment of the Titans by the model.

Last week that would have treated me well, but the situation is an unfavorable one for Tennessee. Will Levis looked good against Atlanta, but a short week on the road against a defense like Pittsburgh does not instill confidence on my end. It would seem the model is too high on the performance of Levis, and I would rather be cautious in my approach there with limited data.

So, that brings in Seattle as the fifth and final play of the week, barring any line changes. Overall I cannot complain about what I see for the contest card this week. There are no outlier edges, and the overall projections are right in line with the betting market, save for a few games.

Last week was a learning experience, and hopefully what I learned leads to better results this weekend.