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.
In week 4, I finally gained a game on Vegas, as my model correctly called Minnesota over Houston and finished the week 11 for 15. On the year, that puts me 3 games back of the betting lines with a 65% accuracy rate. That’s about 1 game behind of where I expected to be at this point.
Week 5 shows the most disparity between my model and Vegas yet. We only agree on 13 of 15 predicted winners, although at the time of this writing, the TEN game looks likely to be postponed, so maybe it is that we agree on 13 of 14.
If so, that’s probably lucky for me as the model likes TEN over BUF, which is a pick I would not consciously make. I also disagree with my model picking CHI over TB, but since Tom Brady has thrown a pick-6 in 3 consecutive games, it is really depressing his numbers and the Chicago pass defense is the real deal. Maybe the model knows something I don’t.
The average difference in spreads from my model and Vegas this week is 3.4 points, which is easily the biggest discrepancy on the year. One of us is way off. I’m pretty sure I know who.
The season expected wins is ever so slowly edging higher, up to 9.8 games. As the season progresses, the played game probabilities are “frozen” and so the season total gets harder to budge.
I said at the beginning of the year, that Rivers would put up his 3 year average efficiency of 0.16 which would translate to 10.5 wins. Today he stands at 0.124 EPA/db so he’ll have to boost his play to get there.