Here’s What An NFL Predictive Modeler Is Thinking About This NFL Sunday

59

One advantage of using a model is the model tells me what to expect throughout the game. Thus, I can generally tell very early on if the game is falling in line with the model projections by using more than the score alone. I can tell by the game plan early in the first quarter. This allows me the opportunity to double down or hedge my way out of positions through in-game betting.

A note on game plans, good teams try to score early, not simply to take the lead but to take the other team out of its game plan and make them a predictable offense. If the team’s game plan does not match what the model thought they should do to be effective, then there is less reason to trust the model. With that in mind, I thought it would be helpful to single out a few select games and reduce them to two questions to show what the model expects. 

Washington v. Buffalo

The model sided with Washington +9 but still liked it at +7.5. 

  1. Is Josh Allen throwing inaccurate balls?
  2. Is Washington’s front four getting pressure without blitzing?

If the answer to both questions above is yes, then Washington is likely to win the game. If the answer to both questions is no, then Washington is likely to lose the game convincingly. If the questions split, then it is a ball game and should end within the neighborhood of the spread.

Tampa Bay v. Los Angeles 

The model sided with Tampa Bay at -1.

  1. Are the Buccaneers passing on first down?
  2. Are the Rams scoring touchdowns only from outside the red zone?

If both answers are yes, the slight edge bends toward the defending champs. If both answers are no, then the Rams have a decided advantage. If the answers split, then it is anybody’s game. 

Bonus Question: Did Joe Buck leap from the third rope and give Deshaun Watson a finishing move?  

New Orleans v. New England

The model passed on this game but had a heavy under lean, a Saints lean, and two player-props. 

  1. Is New Orleans stopping the New England ground game?
  2. Is New Orleans running the ball successfully?

If the answer to both questions is yes, the Saints are in a position to win (and cover, of course), the under is solid, and the player props are live. If both answers are no, it will be another long day in a long season for New Orleans. If the answers split, then the spread should be in play. 

Los Angeles v. Kansas City

The model liked the Chargers plus seven but wanted no part of +6.5.

  1. Are the Chargers handcuffing Herbert on early downs by running?
  2. Is Kansas City starting fast?

If the answer to both questions is yes, the Chargers are toast. If the answer to both questions is no, the Chargers will be in the game to the end. If the answers split, this should fall very close to the line.

Seattle v. Minnesota

The model did not take a side in this game. 

  1. Did Pete Carroll meddle in the offense (run, run, pass, punt?)
  2. Are the Vikings allowing Cousins to spin the ball?

If both answers are yes, then the Vikings are poised for the (slight) upset. If both answers are no, then Seattle should cover easily. If it is a mixed bag, then the game is likely boring and it will come down to the wire.

Chicago v. Cleveland

The model had no side in this game.

  1. Is Nagy coaching the Bears?
  2. Is Nagy coaching the Bears?

If the answer to both questions is yes, Fields will struggle, and the Bears will lose. Can you believe he admitted he had seen a new side of Fields this week because he did not get first-team reps in the preseason!?!?! Isn’t that a firable offense? Who is this guy, and how is he employed?

Nota Bene 

In the preseason, I wrote an article for BLS to point out that while we spend a significant amount of time concerned about the key numbers, mainly three and seven, more games are decided by 17 points or more each season than are determined by three and seven combined. 

Historically, between 23 and 24% of games are decided by three or seven points each season. A whopping 26.99% of games are decided by 17 or more points each season. This year has been no different. Through 33 games, we have the following margin of victory data on these three numbers:

Margin of VictoryNumber of GamesPercentage of Games
Either 3 or 7618.18%
17 or more 824.24%

This is not unusual. It is right on pace. The NFL is a blowout league. 

Nota Bene^2

I was just forced to sit through Trouble With The Curve:

The average life amounts to only 692,000 hours. That movie just robbed me 1.85 of those hours, 0.000002% of my life.

About the author:

+ posts

I write about data and sports. I created my first model in 1997 using nothing more than Excel. Currently, I have data-driven models for the NFL, NBA, and World Cup Soccer.

Mathematics is the music of reason.
— James Joseph Sylvester, English mathematician

Leave a Reply