With week 1 and its complete lack of information out of the way, we’re on to week 2. There isn’t a lot of information about each team per se, but there’s some, and that means I can start running my models. If you need a refresher, click on that link in the banner. But in short: Luigi is my workhorse regression model while Yoshi 1 and Yoshi 2 are young upstarts trying to prove that regularized regression can do better. Each are actually three models (one for picking a winner, one for picking the point difference, and one for picking the total points), but I put them under one name because they use the same predictors and the same method. I always feel a little uncomfortable running them this early, since it feels like the numbers from one game can’t be that great, but I haven’t seen a big fall-off compared to later weeks so far. So off we go!
One game is already in the books; the Ravens had an impressive win over the Steelers. While the models disagreed on who should have won, they agreed on one thing: there should have been more points. Here are the total point predictions, which are usually pretty bad in terms of picking the over/under. But they might be useful for predicting high-scoring, entertaining games, in which case you want to look for Bengals-Falcons and maybe Saints-Browns (that one probably depends on how much you like the Cleveland offense as opposed to the Steelers defense).
|home||away||Over/Under||Luigi||Yoshi 1||Yoshi 2|
As mentioned, the Ravens won already, which is a point to Yoshi 1. But I expect Luigi to come around, particularly since he predicts the Lions will win. Remember here that the numbers for the models are their predicted probability of the home team winning.
|home||away||Home Line||Away Line||Luigi||Yoshi 1||Yoshi 2|
And finally we have the spread picks. Bill Simmons had a whole thing calling this the “year of the ‘dog'”, and then he took each home underdog this week… except for the one playing the Pats. Typical. Yoshi 1, on the other hand, truly does seem to favor the home team in most circumstances.
|home||away||Spread||Odds||Luigi||Yoshi 1||Yoshi 2|
Speaking of Simmons, here are the picks against his lines. His numbers have traditionally been easier to beat than the SBR lines I use for the table above, perhaps because he gets them early in the week for his post. For example, both he and the SuperContest have Minnesota at +3, but SBR shows that the line has moved all the way to +6 now. Obviously a harder sell if you want to take the Patriots to bounce back. He started off well in week 1; we’ll see if it continues.
|home||away||Simmons Spread||Simmons||Luigi||Yoshi 1||Yoshi 2|
Just to wrap up, I need to make some SuperContest picks. If you don’t remember, I pick five games based on how far the model prediction is from the Hilton line; the further away, the better the pick (in theory). And I make five picks for each model, although Luigi and Yoshi 2 often pick the same games. Thanks to their obliteration of the Rams, and a lackluster showing against the Dolphins, the models have all installed the Vikings as solid favorites over the Patriots. Since New England is actually the Vegas/online favorite (and increasingly so given Peterson’s legal trouble), the Vikes are going in the SuperContest. This seems like a good place to point out that I follow the models blindly for these posts; I might pass on the Vikes if picking in real life (last year I would have passed on a number of Packers games due to Rodgers injury issues), but here we’re just interested in what the models say. So Luigi is going with Minnesota +3, Jets +8.5, Lions +3, Titans -3.5, and Steelers +2.5 just edged out Bucs -5.5, so Luigi is starting at 0-1. By a tight margin, Yoshi 2 has the same picks but swaps out the Steelers for Browns +6.5. Yoshi 2 is going with Minnesota, Cleveland, San Diego +6, Kansas City +13, and Oakland +3. We’ll see how it goes!