NFL Lines Model

What is this?

Below are the predictions from my machine learning model. It’s been trained on data from 2009-2018, and has proven to be more accurate than Vegas over time. There are going to be some good weeks, and there are going to be some bad weeks. But over time, these predictions will beat Vegas.

Important before using:

The model doesn’t know how teams rosters change during the offseason, or after trades during the season. So, for example, it has no idea that AB is on Oakland, and that Bell is on the Jets. It’s going to learn this as the season unfolds, but it doesn’t know to start the season.

If there is a pick early in the year (perhaps the Jets and Cardinals in week 1?) that seems wrong based on offseason moves, then I’d take note of that and adjust accordingly. I’ve analyzed previous years and early weeks are just as accurate as late weeks, so this isn’t really a big issue. It’s important to know though.


Spread: Projected spread for the game. This is from the perspective of the home team. A negative number here means the home team is favored by this amount. A positive number means the home team is projected to lose by this amount

Total: Projected total for the game

Home Team Implied Total: Projected point total for the home team

Away Team Implied Total: Projected point total for the opposing team

Week 1 Predictions

Home Team Away Team Predicted Winner Spread Total Home Team Implied Total Away Team Implied Total
CHI GB CHI -5.42 43.71 24.56 19.14
CAR LA LA 1.29 47.37 23.04 24.33
CLE TEN CLE -2.98 45.56 24.27 21.29
JAX KC KC 3.81 44.91 20.55 24.36
MIA BAL BAL 4.4 40.43 18.02 22.42
MIN ATL MIN -1.97 46.71 24.34 22.37
NYJ BUF BUF 2.04 40.25 19.1 21.14
PHI WAS PHI -7.06 41.46 24.26 17.2
LAC IND LAC -1.5 47.27 24.38 22.88
SEA CIN SEA -6.44 41.52 23.98 17.54
ARI DET DET 3.62 40.42 18.4 22.02
DAL NYG DAL -3.95 44.24 24.1 20.14
TB SF TB -1.08 46.05 23.56 22.48
NE PIT NE -5.1 48.9 27 21.9
NO HOU NO -4.8 45.41 25.1 20.3
OAK DEN OAK -2.68 42.01 22.34 19.66