After a quiet summer, the Skeptic celebrates the return of the NFL. Following tradition, I’ll start with predicting who will win each game in week 1. That starts on Thursday! It’s so soon! If you’d like to check on previous iterations of the week 1 picks, here’s the link for last year, which has links for earlier seasons. Basically, I remind you that each year in the NFL is essentially separate from the previous year and it’s tough to make predictions that carry over from one to the other. However, I gallantly do so anyway.
How do I do that? I take the point differential for each team and use it in a regression to predict the week 1 game the following season. For example, in 2011 the Lions outscored their opponents by a total of 87 points over the course of the season (ah, memories). The Rams were outscored by 214. Last year when they played each other in week 1 I would have taken the Lions to win, and so they did. Looking back, the predictions were 9-6 last year (tossing out the Oakland-San Diego game, which the model had right at 50/50). So what do we see happening this year? The table below has the home and away team, last year’s point differential, and the probability of the home team winning.
|home||away||home dif||away dif||home win prob|
|San Francisco||Green Bay||124||97||66%|
To make the table more concrete, let’s look at Thursday night’s game. Denver is hosting Baltimore in a rematch of last year’s great divisional round game. For those of you who already forgot because the Ravens won it all, the Broncos pretty much had the game in hand before allowing the Ravens to tie it and then losing in double overtime. Fitting with the idea that the Broncos were also pretty good last year, they absolutely owned the Ravens in terms of point differential. Denver outscored their opponents by 192 points, or 12 points a game, while the Ravens ‘only’ had a differential of 54, outscoring their opponents by a little over 3.
Given those two numbers (and all the model cares about is 192 and 54), the regression tries to figure out who should win. It also uses home field advantage, which is pretty significant. If two teams were evenly matched, the model would predict the home team to win almost 61% of the time. And if you glance through the predictions for week 1, you can see that home field can make up for fairly large differences in (last year’s) team quality. The Lions were almost 100 points worse than the Vikings last year, but they are slight favorites to win; the same goes for the Cowboys against the Giants. All in all, the only home team that is strongly favored to lose this weekend is Buffalo, although another 8 home teams are essentially toss-ups.
I don’t recommend using these predictions for picking against the spread, because what kind of degenerate uses win probabilities for spread bets?, but if you wanted to make a few choices: according to Bovada right now, your home underdogs in week 1 are San Diego, Carolina, Jacksonville, and the Jets (Jacksonville and the Jets getting 3.5, San Diego and Carolina getting 3). Jacksonville is only at 49% in the chart above, but why not give them a look along with the other three? And in the interest of picking five games for the Hilton SuperContest, let’s add in Cleveland, who is only a 1 point favorite.
Here’s looking forward to another NFL season, where hopefully not too many players are concussed or have their ACLs blown out, and here’s looking forward to hopefully some decent content on the site. Maybe the predictions will even do well this year. Football!