Last week I made some picks for week 1 of the NFL season. Week 1 is a harsh mistress; you have virtually no information about how teams will play. Sure, we have some guesses, but the NFL has so much turnover in terms of players, coaches, and styles to go with the general noise of a single game that it’s pretty tough, in general, to know how things are going to look. But we carry on regardless, because it’s expected of us (I don’t know who expects it, but I’m sure someone does). Here’s how we did:
Broadly speaking, home teams do pretty well in week 1. I expected the home team to win 11 of the 16 games, and the probability was close for most of the others (Tampa was the only team expected to be outclassed. Oops). Overall the model went 9-7, assuming you count the 50/50 call on the Oakland-SD game as a loss. Not as good a job as previous seasons, but what can you do? Over three years now, the model is 32-16 at picking week 1 winners.
I don’t generally recommend it, but for the purposes of comparison I also use the winner picks as picks against the spread. Comparison to who? The indomitable Bill Simmons, of course. His picks are here, and for this week I’ll just use his spread. Both Bill and the model started 0-1 after Dallas beat the Giants (and covered), which wasn’t promising. From there, Bill went 8-7 for an overall week 1 record of 8-8. My week 1 model went 7-8 (I’m giving myself the Chargers since you can’t really expect a 50/50 winner to cover any kind of spread) for an overall record of 7-9. Not a good start, but I also told you not to use winner picks for the spread. I told you!
The other comparison I’m going for is the Hilton SuperContest, in which you pick five games against the spread. You get 2 points for a win, 1 for a push, and 0 for a loss. I went with Cleveland, Tennessee, Arizona, Carolina, and the Jets against the Hilton’s lines. I came through with Cleveland, Arizona, and the Jets, so I’m 3-2 for 6 points. An ok start, I suppose.
Later in the week: week 2 picks using my ‘for real’ models!