Categories

Sunday, August 7, 2016

Effect of New Extra Point Rules on NFL Point Spreads

Last season's extra point rule change saw extra point accuracy decrease from 99.3% in 2014 to 94.2%. This drop is significant, and it did play a role in some games: most notably the AFC championship game.

My question though is whether it affected the public's perception of the expected margin of victory at the end of games, i.e. the gambling spread. Historically, spreads around a multiple of 7 or 3 are very sticky (toggle numbers around these multiples in SBR's half-point calculator and see the difference). In other words, once a spread is around, say, -7, moving to -6.5 or -7.5 is a much bigger jump (percentage-wise) than -5 moving to -4.5 or -5.5. It is much more likely a score finishes at exactly -7, which would greatly increase the chances of one side covering in this example at -6.5 or -7.5.

So did spreads move off of these multiples of 3 or 7 as the season progressed and the public (theoretically) adapted to the lower accuracy on extra point (XP) attempts? I first looked at the closing lines from last season that were on multiples of 7 only throughout the regular season. If XP accuracy decreasing had an effect, we would expect the frequency of spreads to decline throughout the season:


Likewise, we would expect the number of spreads within a half-point of 7 to increase throughout the season:



Neither frequency shows much of a linear trend from the beginning to the end of the season. So I expanded to multiples of both 3 and 7:



As before, we would expect lines exactly on these multiples to decline over the course of the season. And the lines within a half-point to increase:

We actually do see a steady increase from halfway through the season to the end, which supports the hypothesis. But is this enough to conclude anything? Take a look at Week 1. There were 8 spreads within a half-point of a multiple of 3 or 7 (which ended up being the highest number all season) and this was before the "public" had the chance to see how the rule change would affect accuracy (other than preseason data). It's more likely that the trend from Week 9 to Week 16 was simply noise, and not a tangible effect of the 5% difference in XP accuracy.

No comments:

Post a Comment