I really need a keeper, someone who just follows me around and documents everything I do each day so I can remember the damn process when I have to do it again 12 months later.
When I set up the Week 1 S&P+ and F/+ picks last week — actually, when I set up the S&P+ ratings themselves -- I forgot a step in my normal process. The preseason ratings were based on an end-of-season range. As you see, at the end of 2015, Alabama was No. 1 with a +30.0 rating, and North Texas was last at -23.0. So when I set up the distribution for this year’s ratings, I used a range close to that.
One problem: At the beginning of the season, the range is smaller because less is known and teams haven’t separated themselves. The standard deviation is smaller. What constitutes the 99th percentile in Week 1 or 3 is different than in Week 15.
This dawned on me when I checked how the picks were doing on Saturday evening — they were bombing. They were hovering around 42%, and Week 1 is typically a good week for these picks. So I looked into it and found that the games with large projected margins were doing horribly. Like, 30%-against-the-spread horrible. The more tightly projected games were doing just fine. Then what I had done wrong hit me.
So on the running Google Sheet that I’ll be using each week, you’ll find two season totals moving forward: the “Bill screwed up Week 1” total and the “This is what it was supposed to be” total. I tend to post S&P+ (and F/+, which had the same issue) picks simply so people can see that it’s aligned with reality -- I’m not using them to make any money or compete in any pool. So all I care about is what the picks are supposed to say. But if I shared only that moving forward, I would feel like I was cheating. So I’m sharing both.
Week 2 picks (and the Week 2 ratings at FO) are using the proper range. (Among other things, this means the underdog is picked in nearly every blowout spread.) They’ll probably bomb now just to spite me.
- Week 1 S&P+ picks (“Bill screwed up” edition): 16-24-1 (40%)
- Week 1 F/+ picks (“Bill screwed up” edition): 17-23-1 (42%)
- Week 1 S&P+ picks (“how it was supposed to be” edition): 21-19-1 (52%)
- Week 1 F/+ picks (“how it was supposed to be” edition): 22-18-1 (54%)
Week 2 picks
Lots of differences between F/+ and S&P+ this week. That usually works out in F/+’s favor, but we’ll see. (Here’s your reminder that F/+ is the combination of S&P+ and Brian Fremeau’s FEI.)