This is an exercise in building a simple blind predictive model based on a team’s historical stats in 4 areas: pass and rush efficiency, and avg EPA for special teams and penalties. The season long experiment will track how well future games are determined by past statistical performance and compare that record against Vegas betting lines.
Week 7 was another good week. The model again went 11 for 14 and has now pushed ahead of Vegas in the predicted wins race 73 (69.5%) to 72 (68.6%) .
It predicted 2 upsets and delivered on both of them (ARI over SEA, SF over NE). This should have given me a 2 game edge on Vegas, but the lines moved in the DAL-WAS game to favor WAS and I had to give a game back.
Against the spread, the model continues to impress going 9 for 14 (64.3%) and 56.3% on the year, which is a win rate that many professional gamblers would envy. Life has taught me that when you are doing well, nothing bad will ever happen and all your plans will work out perfectly, so I expect this trend to at least continue if not get better.
I want to draw your attention to the 0.3 net points accuracy. That means that relative to actual game point differences, my predicted spreads were off by an average of 0.3 per game. That is ridiculously good and much better than Vegas at 1.7 points per game. It is this accuracy that is allowing me to beat the spread more than 50% of the time . . . so far.
My model is a true believer in Atlanta picking them to win, just like it has in 5 previous weeks, being wrong 4 out of 5 times. It also is picking Seattle to lose, which it has done 4 other times being wrong on 3 of those. Isn’t this supposed to be machine learning?
If I am wrong on these upsets, then Vegas will re-take the lead. I know you are on the edge of your seat. These are early lines and will change before kick-off but I don’t foresee any of them flipping.
Even though the Colts had a week off, the future game probabilities were impacted by week 7.
The expected wins ticked up a bit to 9.4 and the individual game predictions have added some losses dropping that number to a more reasonable 11 wins. 4 of those losses are predicted to come in the next 5 weeks.