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.
Each week, I have been waiting for the model to nosedive and week 11 was pretty close. It predicted 3 upsets and was wrong on every one of them guessing only 6 of 14 winners. In addition it only beat the spread on 6 of 14 games, which isn’t terrible but it’s worse than a coin flip.
Losing 3 games to Vegas puts them back in the lead on the season by 2 games.
While the week was a disappointment, the season numbers are basically right on track with what the model predicted. The goal was to predict 65.7% of games straight up and it is about 2 games ahead of that pace. Against the spread, I expect about 52% and again, it is about 2 games up on that.
The model is picking 3 upsets again this week. TEN over IND, I can sort of see, but the other 2 seem strange to me. I updated my code last week so maybe I’ve got a math issue or maybe that’s just the way it sees things.
Even though the Colts beat green Bay, the expected win totals dropped for the year, sitting at 9.5. The model predicts close games with TEN and PIT and the rest decidedly falling in Indy’s favor.