Prospective 2009 NFL Lines

Earlier this week, SportsBetting released "prospective" lines for each game of the upcoming NFL season. Some sites release "Game of the Year" lines -- just a few games per week, if that -- but S'Bett went all out, offering all 250 post-Week 1 games. When I first saw this, my assumption was that these lines would be reasonably accurate, and would be helpful in projecting the NFL season. When I actually looked at them though, it quickly became obvious that I had overestimated the people that came up with them.

Using these conversions, I went through and made each line into a winning percentage for both team in that game. Adding these up gives us an estimated win total for the full season, something directly comparable with the Football Outsiders win projections. The win totals based on SportsBetting's lines can be found here. Here are the biggest discrepancies between the two sets of projections:

No big surprises there, I don't think. Building on this, we can look at each game and see where FO and S'Bett disagree the most. For example, if FO has the Browns winning three more games than S'Bett, and S'Bett has the Pats winning two more games than FO, then that's a five win discrepancy when those two squads match up. Here are the 10 largest disagreements (team FO likes ATS is underlined, "Diff" is the combined difference between the S'Bett and FO projections for the two teams):

Not suprisingly, since the top three teams in the previous chart were all NFC West teams, that division is prominently featured here. It's definitely easy to look and these and say, "Well, the 49ers won seven games last year, and the Rams won two, so it makes sense that the lines have SF as about six points better". And, as you can see below, last year's win total is definitely relevant to 2009 performance (data below taken from 2006-2008 projections):

But that's not all the information we have. When we add in the FO win projections, they render last year's win total entirely meaningless:

How many games a team won in the previous year doesn't add any information at all to the FO projections; they of course know each team's '08 win total, and appear to weight it perfectly. Which is good to know, since we can easily compare the correlation between S'Bett win total and '08 win total vs. the correlation between FO win total and '08 win total. As it turns out, there's a huge difference; the S'Bett/'08 correlation is 0.759, while the FO/'08 correlation is 0.354.

In English, this means that these lines SportsBetting put out put WAY to much emphasis on how many games a team won last year. So at first glance they tend to make sense, but when you actually consider the advanced stats, something like NO -8 @ STL looks rather absurd. It'll be very interesting to come back to these throughout the season, and see how they change; I would expect that on average they'll move pretty significantly toward the FO projections (so, something like NO -3 @ STL), but we'll have to wait and see.

All 256 lines posted by SportsBetting can be found here.