NFL Week 15 Results

After getting the Cardinals’ pick right on Thursday, things took a turn for the worse with the model picks.  Fortunately, the Lions’ victory has them limping toward the playoffs with somewhat surprising fortitude.  I was happy that the models were right about the Vikings’ cover but Detroit won anyway.  Let’s get through the rest of the results, even if there might not be too many correct picks after those two.

Bill Simmons’s lines resulted in two pushes: Colts-Texans and Seahawks-49ers.  So Bill went 8-6-2 for a season record of 117-102-5.  The models, as has been typical this year, had the worse of it.  Luigi went 5-9-2 for an overall record of 103-116-5, Yoshi 1 went 6-8-2 for a total of 99-120-5, and Yoshi 2 went 5-9-2 for an overall record of 97-122-5.  With two weeks to go, I think Bill has this wrapped up.  I’ll have to see if I can beat him in the playoffs.

Using the SBR lines, Luigi went 4-9-1 for an overall record of 79-82-6.  This is a good time to remind people of (or point out for newer readers) two things.  First, it matters where you get your lines.  I’m not sure where Bill gets his; I look at the range across the SBR website and pick what I think is most common/representative.  Even small differences lead to big changes in overall performance; this week it’s worse for Luigi with SBR because he lost the Seattle push when I had a line of -9.5 instead of 10.  Second, cutting out some close calls matters.  Assume that you bet $110 to win $100 on every game.  With Bill’s lines, picking every game, Luigi has averaged a loss of $8.36 per game.  With the SBR lines, which previous years have told me are tougher than Bill’s lines, but ignoring some games Luigi has averaged a loss of only $6.71 per game.  That’s a big difference in the right direction against harder lines, even if the overall performance has been bad.

Wrapping up the SBR part, Yoshi 1 went 5-9-1 for a season record of 71-92-6 and Yoshi 2 went 4-10-1 for an overall record of 71-93-6.  The consensus picks went 4-8-1 for a total record of 47-59-5.

Luigi’s SuperContest picks were 1-3-1 (got San Fran back as a push) for an overall record of 33-41-1.  Yoshi 1 went 2-2-1 for a total of 32-41-2, and Yoshi 2 went 1-3-1 with the same picks as Luigi for a total of 36-38-1.

This entry was posted in Uncategorized and tagged , , . Bookmark the permalink.

2 Responses to NFL Week 15 Results

  1. Ton says:

    Why do you think the models struggled this year after being so solid the past 2?

    • Alex says:

      I’ve been thinking about this a lot. It could just be chance. Let’s say the model is right around ‘Vegas even’, 53% right. With 256 picks in a season, over enough seasons, sometimes it’ll end up only being right 40% of the time just by bad luck. That’s the optimistic view.

      The pessimistic view is the model somehow became broken. I haven’t changed my code, so it could be bad data scraping (although I try to check that) leading to garbage in garbage out. It could be the the league has changed enough that the old games I have as input (back to 2004) aren’t representative and are throwing off the results. Those would be my top two guesses.

      The fixes for those are to double-check the data, which I’ve done in spots but not thoroughly, and to weigh games so that more recent seasons/this season count more in the model. I’m going to look at that and maybe a more ambitious overhaul in the off season. But I’m also open to other suggestions.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s