2019 NFL Model Preview

With the NFL season quickly approaching, I thought I’d give a little update on my plans for my NFL models this season. Before I talk about this season, though, I want to summarize model performance last season:

Record Win % Profit ROI
Drive Model 60-48-2 .545 6.55u

5.95%

Metric Model

65-43-2 .591 16.09u 14.63%
QBR Model 54-54-2 .491 -4.91u

-4.46%

Consensus

29-17-1 .617 9.36u

19.91%

(for explanations of the models, see the bottom of this post)

While the consensus picks did the best as far as win % and ROI, the volume of the metric model (picked every game) made it more profitable over the course of the season. It’s also important to keep in mind that I only ran the models weeks 10-17, where statistics and metrics were solidified and gave a better representation of team performance. 

As expected, the models performed very poorly in the last two week of the season. Because they only use metrics and statistics, they cannot factor in things such as motivation, injuries, and benching decisions. Here’s a look at the models after week 15, before these factors are exceedingly important.

Record Win % Profit ROI
Drive Model 56-32-1 .629 18.91u

21.25%

Metric Model

58-30-1 .651 22.73u 25.54%
QBR Model 46-42-1 .517 -0.18u

-0.02%

Consensus

26-9-1 .722 14.64u

40.67%

With the poor performance of the QBR model, I took a look at what results would look like if consensus picks only considered the drive and metric models:

Record

Win % Profit ROI
42-21-1 .656 17.18u

26.84%

While the ROI and win % decrease, the overall profit increases due to more volume.

For this season, all my models will remain the same as far as the process goes, and the consensus picks will continue to be the focus. However, because the metrics I use are extremely volatile with small sample sizes, I will not start posting picks until week 4. Hopefully three weeks will be sufficient enough to stabilize data so the models can run how they have in the past. Additionally, I will not post picks for weeks 16 and 17, for the reasons mentioned earlier in the post. 

Other than those two minor changes, I will continue posting the picks as usual on Wednesdays every week. Looking forward to another great season!

 

Model Descriptions:

Drive Model: Uses average drive statistics such as Points per Drive and TD%
Metric Model: Uses advanced metrics such as SRS and EXP
QBR Model: Uses win statistics and team QB Ratings
Consensus Picks: Games in which all the models agree on the same side

Advertisements

3 thoughts on “2019 NFL Model Preview

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s