As always, just go back a post for the predictions (here are week 2’s). In the past my review posts have been mammoth things with all the different predictions and models to track. I’m going to try to pare it down a bit to the most relevant stuff: spread picks, SuperContest picks, and Simmons picks.
Remember that for week 1, the models have no info to go from and so they don’t make any predictions. But I still make some predictions using a different model that just tries to pick a winner, and then I apply those to Bill Simmons’ lines. That didn’t work out too well this year, as I went 6-10 (despite getting a fairly typical 10-6 winner record) while Bill went 9-7. But I’m going to carry that 6-10 over to each model, and they’ll just have to do better than Bill the rest of the way (which they usually do). By my count, Bill was 8-8 this week for a two-week total of 17-15. Luigi was 9-7, Yoshi 1 was 8-8, and Yoshi 2 was 8-8, so Luigi is 15-17 on the year and the Yoshis are at 14-18.
Using the SBR spreads, and ignoring games where the prediction was within a point of the spread (I use this as a rough way to get rid of close calls that could go either way), Luigi was 8-6, Yoshi 1 was 8-5, and Yoshi 2 was 8-6. Not a bad start to the season. There were 8 games where all three models agreed, and they were 5-3 in those games.
Cutting it down even further to the top five games, I make picks against the Hilton lines to play along with their SuperContest. In week 1 I went 4-1 using my alternative model. This week, Luigi went 1-4, getting only the Jets and missing on the Lions, Vikes, Titans, and Steelers. Not so hot, and putting Luigi at 5-5. Yoshi 2 usually takes the same games but this time swapped the Browns for the Steelers, and thus went 2-3 for a 6-4 record. Yoshi 1 likes to go his own way and picked the Vikes, Browns, Chargers, Chiefs, and Raiders. Those were good for 3-2, and so a 7-3 record. The single most popular pick by actual players was the Patriots, showing that sometimes people know more than the model (like that Peterson didn’t play and the line moved three points away from what the SuperContest had it at). Of course, people aren’t perfect; the Lions and 49ers were also popular picks. It’s a long season, and the models usually end up on the fringe of the payline, so there should be more 4-1 weeks in the future.