Welcome back! It is now the start of the season for my NFL spread model. Since I use a 4 week rolling average of each team’s stats, I don’t boot this bad boy up until week 5. I could have used last seasons stats, but I believe teams change too much in the offseason for those to be relevant.
With that said, there are a few updates this year:
- Game changing indicators
- The model now looks at which team is at home, if the two teams are in the same division, and if the team is a “crowd favorite”
- Normalized inputs are better than normal numbers
- The model is now a logarithmic model instead of linear model, since this will normalize the data for each individual input
- If one equation is good, 5 equations is 5x gooder!
- The model was just one equation. Input numbers into a big formula, it spits out a single ‘predicted score’
- New model had 5 equations and combines the outputs to make the pick
Will any of this help? Who knows? But I am excited to find out. Stay tuned!