Kevin Cole's 2018 NFL Betting Model: Super Bowl
Along with working extensively on NFL projections this season, I’m also putting the data to use projecting game outcomes.
Last week, we introduced a new model for side bets only and the model’s picks will only once during the week. The first version of the model is what I will call 1.0, and this model is the results of a lot of work during the season to develop a power-ranking system based on play-by-play data.
The data I’m using also has advanced metrics, including expected points added (EPA) and success rates based on EPA.
I will continue track total record for the year, along with what the results are of the new model.
Methodology
The betting model is based on a combination of offense and defensive play-by-play stats, including success rate, or how often a play has positive EPA, yards per play, and EPA per play. The offensive and defense metrics are calculated for every team on a weekly basis and those trailing results are the basis for projecting team strength in upcoming matchups.
Conference Championship Week Review
Last week the model was 1-0 on plays, while leans were 0-1.
For full disclosure, here are the numbers for the new and old model that has been discontinued.
Conference Championship Picks
Picks as of Saturday at roughly 10 a.m. ET
These picks are based on the power rankings above, with adjustments for home field advantage, and additional/fewer days of rest. The darker green/red plays are the plays for home/away team on the week. The lighter green/red cells are the leans of the week.
Plays:
None
Leans:
Rams +2.5
