As we get closer to the end of the season, there’s more pressure (at least mentally, for me) on the models to have good weeks in an effort to get to something in the neighborhood of .500. I usually start writing these summaries without taking a close look at the results, but my vague impression is that things went ok this week. Let’s see if I was right.
We finally had a push again with Simmons’ picks, as the Vikings won by 6. So Bill went 9-6-1 this week for an overall record of 109-96-3. Unfortunately, he beat out the models again. Luigi went 8-7-1 for an overall record of 98-107-3, Yoshi 1 went 6-9-1 for an overall record of 93-112-3, and Yoshi 2 went 8-7-1 for an overall record of 92-113-3. Luigi is closest but it behind by 11 games. I don’t want to say that lead is insurmountable with three weeks to go, but it’s insurmountable.
On their own picks Luigi went 8-7 for an overall record of 75-73-5, Yoshi 1 went 7-9 for an overall record of 66-83-5, and Yoshi 2 went 8-7 for an overall record of 67-83-5. The consensus picks went 5-6 for an overall record of 43-51-4.
I have to say I’m most disappointed in Yoshi 1. In previous years there would be weeks where multiple games got the same expected win percentage or point differential – the model wouldn’t converge to any particular weights, so it basically went with home field. That hasn’t really happened this year, so I thought maybe we hit a tipping point in the amount of data (remember that Yoshi 1 uses only the week of interest, so even with 10 years of data there are only 120-130 games for some weeks, and many predictor variables) and the predictions would start coming around. It’s obviously a down year overall, but I hoped Yoshi 1 would weather it better instead of worse.
To wrap up, Luigi’s SuperContest picks went 3-2, Yoshi 1’s went 2-2-1, and Yoshi 2’s went 3-2. Their overall records are 32-38, 30-39-1 (I guess it’s impressive that there weren’t any pushes until now), and 35-35 (.500!).